Skip to content

πŸ₯ 🎀 The largest clinical study in the world to collect voice data labeled with health information (N>6,000 participants, 48 utterances, 80+ health labels). Preprint: https://www.medrxiv.org/content/10.1101/2021.08.16.21262125v1

License

Notifications You must be signed in to change notification settings

jim-schwoebel/voiceome

Repository files navigation

Voiceome

The Voiceome Study is the world’s largest clinical study to collect voice information labeled with health traits. The goal is to recruit between 10,000-100,000 patients into the study and track them at multiple time points to advance scientific understanding for vocal biomarkers and speed up the process for its commercialization.

More information about the Voiceome study can be found @ the wiki and website.

Background/goals

One of the primary causes of slow translation of voice biomarker research is a result of small datasets. There have been many research studies that have shown that acoustic or linguistic voice features relate to health conditions like stress, schizophrenia, depression, bipolar disorder, stroke, and cardiovascular diseases. However, many of these studies have relatively small sample sizes (N<1,000 patients in each class), suffer from poor audio quality (e.g. use of different microphones and data collection techniques), and are difficult to access (e.g. large licensing fees to corporations and/or difficulty obtaining IRB access without close academic ties). These factors dramatically limit the ability to model the underlying data and correlate voice data to health traits, replicate peer-reviewed publications, as well as open up this work to the broader scientific community.

Therefore, the main goals of the Voiceome Study include:

  • Standardizing collection tools and features - Standardize the tools and techniques used to collect and featurize voice samples for use in health-related research. We have done this by building SurveyLex, a web-enabled survey tool for vocal biomarker research.
  • Establishing reference ranges - Create a standard set of normative ranges of acoustic and linguistic features across various ages, genders, microphone types, and dialects labeled with health information. These are provided as tables and standardized scripts with the command line interface in this repository.
  • Standardizing voice protocols - Standardize the series of voice tasks used to collect speech samples for use speech and language biomarker research. These voice protocols can be found in the wiki, and includes novel tasks like automated confrontational naming tasks and nonword tasks.
  • Standardizing health trait labels - Standardize the type of health trait labels tied to voice information to emit the most signal from the voice. These are provided here.
  • Creating machine learning models - From Voiceome Study dataset, build and optimize novel machine learning models; to publish this work in academic journals openly. These models will be published on in follow-on publications into the future.
  • Creating an ethics framework - Create a venue for legal, ethical, and security considerations for collecting voice data for health-related purposes. This can be read in the Voiceome strategic plan.
  • Build a lasting community - The Voiceome project will foster a lasting community and enthusiasm for the field of vocal biomarkers in general. We do this through writing textbooks like Voice Computing in Python, collaborating with research organizations like the University of Washington and Biogen, and through training fellows through the Innovation Fellows Program.

Clinical trial design

Anyone >18 years of age that could speak English was eligible to participate in the Voiceome Study (>6,000 participants completed the first survey). Each participant was randomly assigned one of three tracks (AAAA, ABAB, ABCD, or ACBD) and could participate in follow-on surveys separated by roughly three weeks in these tracks. Subjects were primarily recruited on Amazon Mechanical Turk with a paid incentive ($5-20/session), and anyone who completed 4 surveys was eligible to receive a health report.

Overvew Locations

Getting started

This assumes you are using a MacOS-related computer with a MacOS Big Sur operating system.

MacOS

To get started, clone this repository in a virtual environment and install requirements.txt (note this can take 10 minutes):

git clone [email protected]:jim-schwoebel/voiceome.git
cd voiceome

Now install various libraries with Homebrew (if you don't have Homebrew, install it before this step following these instructions):

brew install ffmpeg sox autoconf automake m4 libtool autoconf gcc portaudio lasound pyenv pyenv-virtualenv bzip2 zlib llvm@8
brew unlink python
brew link bzip2

Now install Xcode if you have not already:

sudo xcode-select --install

Edit some environment variables in .bash_profile:

nano ~/.bash_profile

Then paste this in:

export LDFLAGS="-L/usr/local/opt/llvm@8/lib"
export CPPFLAGS="-I/usr/local/opt/llvm@8/include"
eval "$(pyenv init -)"
export PATH="/usr/local/opt/llvm@8/bin:$PATH"

Now go back to the main terminal by pressing command + x and then press -y to save changes.

Now setup python 3.9.5 with pyenv:

echo 'eval "$(pyenv init -)"' >> ~/.zshrc 
cd /Users/$USER/.pyenv/plugins/python-build/../.. && git pull && cd -
LDFLAGS="-L/usr/local/opt/zlib/lib -L/usr/local/opt/bzip2/lib" \
CPPFLAGS="-I/usr/local/opt/zlib/include -I/usr/local/opt/bzip2/include" \
pyenv install 3.9.5
pyenv global 3.9.5

Now install dependencies (along with matplotlib, new version):

pip3 install --upgrade pip
pip3 install -r requirements.txt
pip3 install https://github.com/matplotlib/matplotlib/archive/refs/tags/v3.4.2.tar.gz

Next, install OpenSMILE:

cd scripts/features/helpers/opensmile/opensmile-2.3.0
bash autogen.sh
bash autogen.sh
./configure
make -j4 ; make
make install

Optionally, if you want to transcribe data with Microsoft Azure (this was the transcriber used in the Voiceome paper), update the 'settings.json' Azure Key:

{"AzureKey": "KEY_GOES_HERE",...}

You can now use the convenient Voiceome CLI tool for a few things:

Usage: cli.py [options]

Options:
  -h, --help            show this help message and exit
  --c=command, --command=command
                        the target command (cleaning API = 'clean',  features
                        API = 'features',  quality API = 'quality',
                        references API = 'reference',  samples API = 'sample',
                        testing API = 'test',  urls API = 'urls',  visualize
                        API = 'visualize',  list/change default settings =
                        'settings')
  --a=age_gender, --agegender=age_gender
                        specify the age and gender in CamelCase for references
                        ('TwentiesMale' =  male aged 20s); if not used will
                        default to settings.json value.
  --d=dir, --dir=dir    an array of the target directory (or directories) that
                        contains sample files all APIs (e.g.
                        '/Users/jim/desktop/allie/train_dir/teens/')
  --e=feature_embedding, --embedding=feature_embedding
                        the feature embedding to use for reference ranges
                        (e.g. 'OpenSmile'); if not used it will default to
                        settings.json value.
  --f=feature, --feature=feature
                        the feature value in a feature embedding to use for a
                        reference range (e.g.
                        'F0semitoneFrom27.5Hz_sma3nz_amean'); if not used it
                        will default to settings.json value.
  --fi=file, --file=file
                        an audio file to extract relevant quality metrics and
                        transcribe (e.g. 'test.wav')
  --t=task, --task=task
                        the task type to focus on (e.g. 'microphone_task'); if
                        not used it will default to settings.json value.
  --v=visualizationtype, --vtype=visualizationtype
                        the visualization type that you'd like to use - two
                        options: ['bar', 'bar_cohorts']
  --verbosity=verbosity
                        whether or not to display visualizations/charts on the
                        screen ([True or False]).
  --u=urls, --urls=urls
                        the url links for surveys in the Voiceome Study
                        (useful for cloning surveys via the SurveyLex
                        interface).

Using the CLI

There are many things you can do with the Voiceome CLI. Primarily, you can use it to extract features and references, as well as make cool visualizations to do quick experiments from the means and standard deviations of the features. A full list of things is below:

If you have any thing else you'd find valuable - feel free to suggest some new features here!

cleaning data

You can clean data with the CLI by specifying the command and directory of interest. This will clean files to mono 16000 Hz:

python3 cli.py --command clean --dir /Users/jimschwoebel/desktop/voiceome/data/test/a871b730-cc8a-11eb-a78c-b9f05e289d42

This will then clean audio files with FFmpeg:

a871b730-cc8a-11eb-a78c-b9f05e289d42:  82%|β–ˆβ–ˆβ–ˆβ–ˆ | 41/50 [00:02<00:00, 18.99it/s]ffmpeg version 4.3.1 Copyright (c) 2000-2020 the FFmpeg developers
  built with Apple clang version 11.0.3 (clang-1103.0.32.62)
  configuration: --prefix=/usr/local/Cellar/ffmpeg/4.3.1 --enable-shared --enable-pthreads --enable-version3 --enable-avresample --cc=clang --host-cflags= --host-ldflags= --enable-ffplay --enable-gnutls --enable-gpl --enable-libaom --enable-libbluray --enable-libdav1d --enable-libmp3lame --enable-libopus --enable-librav1e --enable-librubberband --enable-libsnappy --enable-libsrt --enable-libtesseract --enable-libtheora --enable-libvidstab --enable-libvorbis --enable-libvpx --enable-libwebp --enable-libx264 --enable-libx265 --enable-libxml2 --enable-libxvid --enable-lzma --enable-libfontconfig --enable-libfreetype --enable-frei0r --enable-libass --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenjpeg --enable-librtmp --enable-libspeex --enable-libsoxr --enable-videotoolbox --disable-libjack --disable-indev=jack
  libavutil      56. 51.100 / 56. 51.100
  libavcodec     58. 91.100 / 58. 91.100
  libavformat    58. 45.100 / 58. 45.100
  libavdevice    58. 10.100 / 58. 10.100
  libavfilter     7. 85.100 /  7. 85.100
  libavresample   4.  0.  0 /  4.  0.  0
  libswscale      5.  7.100 /  5.  7.100
  libswresample   3.  7.100 /  3.  7.100
  libpostproc    55.  7.100 / 55.  7.100
Guessed Channel Layout for Input Stream #0.0 : stereo
Input #0, wav, from 'nlx-b558d3b0-cc8b-11eb-aefd-7de9011dbebd.wav':
  Duration: 00:00:01.76, bitrate: 1411 kb/s
    Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 44100 Hz, stereo, s16, 1411 kb/s
Stream mapping:
  Stream #0:0 -> #0:0 (pcm_s16le (native) -> pcm_s16le (native))
Press [q] to stop, [?] for help
Output #0, wav, to 'nlx-b558d3b0-cc8b-11eb-aefd-7de9011dbebd_cleaned.wav':
  Metadata:
    ISFT            : Lavf58.45.100
    Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 16000 Hz, mono, s16, 256 kb/s
    Metadata:
      encoder         : Lavc58.91.100 pcm_s16le
size=      55kB time=00:00:01.76 bitrate= 256.4kbits/s speed= 676x    
video:0kB audio:55kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: 0.138122%
ffmpeg version 4.3.1 Copyright (c) 2000-2020 the FFmpeg developers
  built with Apple clang version 11.0.3 (clang-1103.0.32.62)
  configuration: --prefix=/usr/local/Cellar/ffmpeg/4.3.1 --enable-shared --enable-pthreads --enable-version3 --enable-avresample --cc=clang --host-cflags= --host-ldflags= --enable-ffplay --enable-gnutls --enable-gpl --enable-libaom --enable-libbluray --enable-libdav1d --enable-libmp3lame --enable-libopus --enable-librav1e --enable-librubberband --enable-libsnappy --enable-libsrt --enable-libtesseract --enable-libtheora --enable-libvidstab --enable-libvorbis --enable-libvpx --enable-libwebp --enable-libx264 --enable-libx265 --enable-libxml2 --enable-libxvid --enable-lzma --enable-libfontconfig --enable-libfreetype --enable-frei0r --enable-libass --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenjpeg --enable-librtmp --enable-libspeex --enable-libsoxr --enable-videotoolbox --disable-libjack --disable-indev=jack
  libavutil      56. 51.100 / 56. 51.100
  libavcodec     58. 91.100 / 58. 91.100
  libavformat    58. 45.100 / 58. 45.100
  libavdevice    58. 10.100 / 58. 10.100
  libavfilter     7. 85.100 /  7. 85.100
  libavresample   4.  0.  0 /  4.  0.  0
  libswscale      5.  7.100 /  5.  7.100
  libswresample   3.  7.100 /  3.  7.100
  libpostproc    55.  7.100 / 55.  7.100
Guessed Channel Layout for Input Stream #0.0 : stereo
Input #0, wav, from 'nlx-be1ffd20-cc8b-11eb-aefd-7de9011dbebd.wav':
  Duration: 00:00:01.49, bitrate: 1411 kb/s
    Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 44100 Hz, stereo, s16, 1411 kb/s
Stream mapping:
  Stream #0:0 -> #0:0 (pcm_s16le (native) -> pcm_s16le (native))
Press [q] to stop, [?] for help
Output #0, wav, to 'nlx-be1ffd20-cc8b-11eb-aefd-7de9011dbebd_cleaned.wav':
  Metadata:
    ISFT            : Lavf58.45.100
    Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 16000 Hz, mono, s16, 256 kb/s
    Metadata:
      encoder         : Lavc58.91.100 pcm_s16le
size=      47kB time=00:00:01.48 bitrate= 256.4kbits/s speed= 602x    
video:0kB audio:46kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: 0.164024%
ffmpeg version 4.3.1 Copyright (c) 2000-2020 the FFmpeg developers
  built with Apple clang version 11.0.3 (clang-1103.0.32.62)
  configuration: --prefix=/usr/local/Cellar/ffmpeg/4.3.1 --enable-shared --enable-pthreads --enable-version3 --enable-avresample --cc=clang --host-cflags= --host-ldflags= --enable-ffplay --enable-gnutls --enable-gpl --enable-libaom --enable-libbluray --enable-libdav1d --enable-libmp3lame --enable-libopus --enable-librav1e --enable-librubberband --enable-libsnappy --enable-libsrt --enable-libtesseract --enable-libtheora --enable-libvidstab --enable-libvorbis --enable-libvpx --enable-libwebp --enable-libx264 --enable-libx265 --enable-libxml2 --enable-libxvid --enable-lzma --enable-libfontconfig --enable-libfreetype --enable-frei0r --enable-libass --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenjpeg --enable-librtmp --enable-libspeex --enable-libsoxr --enable-videotoolbox --disable-libjack --disable-indev=jack
  libavutil      56. 51.100 / 56. 51.100
  libavcodec     58. 91.100 / 58. 91.100
  libavformat    58. 45.100 / 58. 45.100
  libavdevice    58. 10.100 / 58. 10.100
  libavfilter     7. 85.100 /  7. 85.100
  libavresample   4.  0.  0 /  4.  0.  0
  libswscale      5.  7.100 /  5.  7.100
  libswresample   3.  7.100 /  3.  7.100
  libpostproc    55.  7.100 / 55.  7.100
Guessed Channel Layout for Input Stream #0.0 : stereo
Input #0, wav, from 'nlx-b9e2a190-cc8b-11eb-aefd-7de9011dbebd.wav':
  Duration: 00:00:02.04, bitrate: 1411 kb/s
    Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 44100 Hz, stereo, s16, 1411 kb/s
Stream mapping:
  Stream #0:0 -> #0:0 (pcm_s16le (native) -> pcm_s16le (native))
Press [q] to stop, [?] for help
Output #0, wav, to 'nlx-b9e2a190-cc8b-11eb-aefd-7de9011dbebd_cleaned.wav':
  Metadata:
    ISFT            : Lavf58.45.100
    Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 16000 Hz, mono, s16, 256 kb/s
    Metadata:
      encoder         : Lavc58.91.100 pcm_s16le
size=      64kB time=00:00:02.04 bitrate= 256.3kbits/s speed= 705x    
video:0kB audio:64kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: 0.119288%
...

featurizing data

You can also featurize a folder of files with:

python3 cli.py --command features --dir /Users/jimschwoebel/desktop/voiceome/data/test/a871b730-cc8a-11eb-a78c-b9f05e289d42

This will then featurize all the files with OpenSmileFeatures, ProsodyFeatures, PauseFeatures, and AudiotextFeatures via the Allie ML framework:

Requirement already satisfied: numba==0.48 in /usr/local/lib/python3.8/site-packages (0.48.0)
Requirement already satisfied: setuptools in /usr/local/lib/python3.8/site-packages (from numba==0.48) (49.2.0)
Requirement already satisfied: numpy>=1.15 in /usr/local/lib/python3.8/site-packages (from numba==0.48) (1.17.3)
Requirement already satisfied: llvmlite<0.32.0,>=0.31.0dev0 in /usr/local/lib/python3.8/site-packages (from numba==0.48) (0.31.0)
a871b730-cc8a-11eb-a78c-b9f05e289d42:   0%|              | 0/50 [00:00<?, ?it/s]deepspeech_dict transcribing: nlx-b66e6260-cc8b-11eb-aefd-7de9011dbebd_cleaned.wav
--2021-06-16 15:45:37--  https://github.com/mozilla/DeepSpeech/releases/download/v0.7.0/deepspeech-0.7.0-models.pbmm
Resolving github.com (github.com)... 140.82.113.3
Connecting to github.com (github.com)|140.82.113.3|:443... connected.
WARNING: cannot verify github.com's certificate, issued by β€˜CN=Bitdefender CA SSL,O=Endpoint,L=Bucharest,ST=Bucharest,C=RO’:
  Self-signed certificate encountered.
HTTP request sent, awaiting response... 302 Found
Location: https://github-releases.githubusercontent.com/60273704/db3b3f80-84bd-11ea-93d7-1ddb76a21efe?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAIWNJYAX4CSVEH53A%2F20210616%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20210616T194537Z&X-Amz-Expires=300&X-Amz-Signature=ac030f0772fcefa16d7d8707dab78a2397d796101946d6090feaf865f50db6b8&X-Amz-SignedHeaders=host&actor_id=0&key_id=0&repo_id=60273704&response-content-disposition=attachment%3B%20filename%3Ddeepspeech-0.7.0-models.pbmm&response-content-type=application%2Foctet-stream [following]
--2021-06-16 15:45:37--  https://github-releases.githubusercontent.com/60273704/db3b3f80-84bd-11ea-93d7-1ddb76a21efe?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAIWNJYAX4CSVEH53A%2F20210616%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20210616T194537Z&X-Amz-Expires=300&X-Amz-Signature=ac030f0772fcefa16d7d8707dab78a2397d796101946d6090feaf865f50db6b8&X-Amz-SignedHeaders=host&actor_id=0&key_id=0&repo_id=60273704&response-content-disposition=attachment%3B%20filename%3Ddeepspeech-0.7.0-models.pbmm&response-content-type=application%2Foctet-stream
Resolving github-releases.githubusercontent.com (github-releases.githubusercontent.com)... 185.199.109.154, 185.199.110.154, 185.199.111.154, ...
Connecting to github-releases.githubusercontent.com (github-releases.githubusercontent.com)|185.199.109.154|:443... connected.
WARNING: cannot verify github-releases.githubusercontent.com's certificate, issued by β€˜CN=Bitdefender CA SSL,O=Endpoint,L=Bucharest,ST=Bucharest,C=RO’:
  Self-signed certificate encountered.
HTTP request sent, awaiting response... 200 OK
Length: 188916323 (180M) [application/octet-stream]
Saving to: β€˜deepspeech-0.7.0-models.pbmm’

ch-0.7.0-models.pbm  53%[=========>          ]  97.11M  25.8MB/s    et
...

extracting quality metrics

You can featurize an audio file and get back quality metrics with:

python3 cli.py --command quality --file test.wav --dir /Users/jimschwoebel/desktop/voiceome/tests

You then get back a list of quality metrics from the Voiceome paper:

--------------------------------------------------------------
___  ___ _____ ___________ _____ _____  _____          
|  \/  ||  ___|_   _| ___ \_   _/  __ \/  ___|         
| .  . || |__   | | | |_/ / | | | /  \/\ `--.   ______ 
| |\/| ||  __|  | | |    /  | | | |     `--. \ |______|
| |  | || |___  | | | |\ \ _| |_| \__/\/\__/ /         
\_|  |_/\____/  \_/ \_| \_|\___/ \____/\____/          
                                                       
                                                       
 _____ _____ _____ _____ _    _  ___  _   _ 
|_   _|  ___/  ___|_   _| |  | |/ _ \| | | |
  | | | |__ \ `--.  | | | |  | / /_\ \ | | |
  | | |  __| `--. \ | | | |/\| |  _  | | | |
  | | | |___/\__/ / | |_\  /\  / | | \ \_/ /
  \_/ \____/\____/  \_(_)\/  \/\_| |_/\___/ 
                                            
                                            

--------------------------------------------------------------
{'brownfox_feature': 7.1022727272727275, 'animal_feature': 0, 'letterf_feature': 4, 'caterpillar_feature': 1.6441573693482088, 'mandog_feature': 10.586552217453505, 'tourbus_feature': 14.686248331108146, 'bnt_feature': 0, 'nonword_feature': 2}

querying references

Feature relative to age and gender

You can get a reference feature with a specified age and gender (e.g. TwentiesMale) with:

python3 cli.py --command reference --vtype table_by_feature --agegender TwentiesMale

Which returns:

{'AverageValue': 24.907, 'StdValue': 5.774, 'Description': 'Twenties Male (Ages 18-29)', 'SampleNumber': 371}

Feature embedding for a speech task

You can get a reference table for a speech task (e.g. microphone task) and feature embedding (e.g. OpensmileFeatures) with:

python3 cli.py --command reference --vtype table_by_embedding --agegender TwentiesMale

Which returns:

+--------+--------+---------------------+---------+----------+---------+-------+
|  Task  | Featur |       Feature       | AgeGend | Average  | Standar | Sampl |
|        | eType  |                     |   er    |          | d Devia | e Num |
|        |        |                     |         |          |  tion   |  ber  |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F0semitoneFrom27.5H | ['Twent |  24.907  |  5.774  |  371  |
| hone_t | ileFea |   z_sma3nz_amean    | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F0semitoneFrom27.5H | ['Twent |  0.159   |  0.106  |  371  |
| hone_t | ileFea | z_sma3nz_stddevNorm | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F0semitoneFrom27.5H | ['Twent |  22.54   |  5.598  |  371  |
| hone_t | ileFea | z_sma3nz_percentile | iesMale |          |         |       |
|  ask   | tures  |        20.0         |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F0semitoneFrom27.5H | ['Twent |  24.146  |  5.716  |  371  |
| hone_t | ileFea | z_sma3nz_percentile | iesMale |          |         |       |
|  ask   | tures  |        50.0         |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F0semitoneFrom27.5H | ['Twent |  26.313  |  6.209  |  371  |
| hone_t | ileFea | z_sma3nz_percentile | iesMale |          |         |       |
|  ask   | tures  |        80.0         |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F0semitoneFrom27.5H | ['Twent |  3.773   |  2.744  |  371  |
| hone_t | ileFea | z_sma3nz_pctlrange0 | iesMale |          |         |       |
|  ask   | tures  |         -2          |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F0semitoneFrom27.5H | ['Twent |  181.24  | 293.03  |  371  |
| hone_t | ileFea | z_sma3nz_meanRising | iesMale |          |         |       |
|  ask   | tures  |        Slope        |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F0semitoneFrom27.5H | ['Twent | 177.741  | 314.493 |  371  |
| hone_t | ileFea | z_sma3nz_stddevRisi | iesMale |          |         |       |
|  ask   | tures  |       ngSlope       |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F0semitoneFrom27.5H | ['Twent |  60.672  | 97.411  |  371  |
| hone_t | ileFea | z_sma3nz_meanFallin | iesMale |          |         |       |
|  ask   | tures  |       gSlope        |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F0semitoneFrom27.5H | ['Twent |  57.899  | 120.13  |  371  |
| hone_t | ileFea | z_sma3nz_stddevFall | iesMale |          |         |       |
|  ask   | tures  |      ingSlope       |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | loudness_sma3_amean | ['Twent |  0.349   |  0.252  |  371  |
| hone_t | ileFea |                     | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | loudness_sma3_stdde | ['Twent |  0.901   |  0.272  |  371  |
| hone_t | ileFea |        vNorm        | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | loudness_sma3_perce | ['Twent |  0.089   |  0.101  |  371  |
| hone_t | ileFea |      ntile20.0      | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | loudness_sma3_perce | ['Twent |  0.229   |  0.191  |  371  |
| hone_t | ileFea |      ntile50.0      | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | loudness_sma3_perce | ['Twent |  0.593   |  0.443  |  371  |
| hone_t | ileFea |      ntile80.0      | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | loudness_sma3_pctlr | ['Twent |  0.504   |  0.398  |  371  |
| hone_t | ileFea |       ange0-2       | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | loudness_sma3_meanR | ['Twent |  6.433   |  4.592  |  371  |
| hone_t | ileFea |     isingSlope      | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | loudness_sma3_stdde | ['Twent |  3.688   |  2.694  |  371  |
| hone_t | ileFea |    vRisingSlope     | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | loudness_sma3_meanF | ['Twent |  4.804   |  3.464  |  371  |
| hone_t | ileFea |     allingSlope     | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | loudness_sma3_stdde | ['Twent |  2.867   |  2.18   |  371  |
| hone_t | ileFea |    vFallingSlope    | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | jitterLocal_sma3nz_ | ['Twent |   0.03   |  0.013  |  371  |
| hone_t | ileFea |        amean        | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | jitterLocal_sma3nz_ | ['Twent |  1.082   |  0.428  |  371  |
| hone_t | ileFea |     stddevNorm      | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | shimmerLocaldB_sma3 | ['Twent |  1.413   |  0.404  |  371  |
| hone_t | ileFea |      nz_amean       | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | shimmerLocaldB_sma3 | ['Twent |  0.802   |  0.21   |  371  |
| hone_t | ileFea |    nz_stddevNorm    | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | HNRdBACF_sma3nz_ame | ['Twent |   4.23   |  1.784  |  371  |
| hone_t | ileFea |         an          | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | HNRdBACF_sma3nz_std | ['Twent |  0.692   |  1.332  |  371  |
| hone_t | ileFea |       devNorm       | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | logRelF0-H1-H2_sma3 | ['Twent |  3.783   |  6.174  |  371  |
| hone_t | ileFea |      nz_amean       | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | logRelF0-H1-H2_sma3 | ['Twent |  -2.468  | 203.594 |  371  |
| hone_t | ileFea |    nz_stddevNorm    | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | logRelF0-H1-A3_sma3 | ['Twent |  22.015  |  8.122  |  371  |
| hone_t | ileFea |      nz_amean       | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | logRelF0-H1-A3_sma3 | ['Twent |  0.511   |  0.662  |  371  |
| hone_t | ileFea |    nz_stddevNorm    | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F1frequency_sma3nz_ | ['Twent |  547.83  | 126.534 |  371  |
| hone_t | ileFea |        amean        | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F1frequency_sma3nz_ | ['Twent |  0.307   |  0.079  |  371  |
| hone_t | ileFea |     stddevNorm      | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F1bandwidth_sma3nz_ | ['Twent | 1406.105 | 246.924 |  371  |
| hone_t | ileFea |        amean        | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F1bandwidth_sma3nz_ | ['Twent |  0.156   |  0.055  |  371  |
| hone_t | ileFea |     stddevNorm      | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F1amplitudeLogRelF0 | ['Twent | -137.468 | 24.801  |  371  |
| hone_t | ileFea |    _sma3nz_amean    | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F1amplitudeLogRelF0 | ['Twent |  -0.643  |  0.213  |  371  |
| hone_t | ileFea | _sma3nz_stddevNorm  | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F2frequency_sma3nz_ | ['Twent | 1443.097 | 255.927 |  371  |
| hone_t | ileFea |        amean        | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F2frequency_sma3nz_ | ['Twent |  0.143   |  0.036  |  371  |
| hone_t | ileFea |     stddevNorm      | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F2amplitudeLogRelF0 | ['Twent | -139.385 | 23.984  |  371  |
| hone_t | ileFea |    _sma3nz_amean    | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F2amplitudeLogRelF0 | ['Twent |   -0.6   |  0.192  |  371  |
| hone_t | ileFea | _sma3nz_stddevNorm  | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F3frequency_sma3nz_ | ['Twent | 2401.48  | 408.194 |  371  |
| hone_t | ileFea |        amean        | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F3frequency_sma3nz_ | ['Twent |  0.081   |  0.023  |  371  |
| hone_t | ileFea |     stddevNorm      | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F3amplitudeLogRelF0 | ['Twent | -141.62  | 23.205  |  371  |
| hone_t | ileFea |    _sma3nz_amean    | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | F3amplitudeLogRelF0 | ['Twent |  -0.565  |  0.178  |  371  |
| hone_t | ileFea | _sma3nz_stddevNorm  | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | alphaRatioV_sma3nz_ | ['Twent | -13.234  |  5.317  |  371  |
| hone_t | ileFea |        amean        | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | alphaRatioV_sma3nz_ | ['Twent |  -0.603  |  0.421  |  371  |
| hone_t | ileFea |     stddevNorm      | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | hammarbergIndexV_sm | ['Twent |  24.245  |  6.874  |  371  |
| hone_t | ileFea |     a3nz_amean      | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | hammarbergIndexV_sm | ['Twent |  0.364   |  0.187  |  371  |
| hone_t | ileFea |   a3nz_stddevNorm   | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | slopeV0-500_sma3nz_ | ['Twent |  0.028   |  0.03   |  371  |
| hone_t | ileFea |        amean        | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | slopeV0-500_sma3nz_ | ['Twent |  0.841   | 10.358  |  371  |
| hone_t | ileFea |     stddevNorm      | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | slopeV500-1500_sma3 | ['Twent |  -0.014  |  0.012  |  371  |
| hone_t | ileFea |      nz_amean       | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | slopeV500-1500_sma3 | ['Twent |  8.402   | 203.814 |  371  |
| hone_t | ileFea |    nz_stddevNorm    | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | alphaRatioUV_sma3nz | ['Twent |  -8.073  |  5.94   |  371  |
| hone_t | ileFea |       _amean        | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | hammarbergIndexUV_s | ['Twent |  17.64   |  6.902  |  371  |
| hone_t | ileFea |     ma3nz_amean     | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | slopeUV0-500_sma3nz | ['Twent |  0.015   |  0.034  |  371  |
| hone_t | ileFea |       _amean        | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | slopeUV500-1500_sma | ['Twent |  0.011   |  0.011  |  371  |
| hone_t | ileFea |      3nz_amean      | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | loudnessPeaksPerSec | ['Twent |  2.904   |  0.901  |  371  |
| hone_t | ileFea |                     | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | VoicedSegmentsPerSe | ['Twent |  1.923   |  0.742  |  371  |
| hone_t | ileFea |          c          | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | MeanVoicedSegmentLe | ['Twent |  0.165   |  0.072  |  371  |
| hone_t | ileFea |       ngthSec       | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | StddevVoicedSegment | ['Twent |  0.132   |  0.077  |  371  |
| hone_t | ileFea |      LengthSec      | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | MeanUnvoicedSegment | ['Twent |  0.506   |  0.996  |  371  |
| hone_t | ileFea |       Length        | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+
| microp | Opensm | StddevUnvoicedSegme | ['Twent |  0.414   |  0.396  |  371  |
| hone_t | ileFea |      ntLength       | iesMale |          |         |       |
|  ask   | tures  |                     |   ']    |          |         |       |
+--------+--------+---------------------+---------+----------+---------+-------+

Feature across speech tasks

You can get a reference table across sample tasks while keeping a feature constant with:

python3 cli.py --command reference --vtype table_across_tasks --agegender TwentiesMale

Which returns:

+--------+---------+------------------+----------+---------+----------+--------+
|  Task  | Feature |     Feature      | AgeGende | Average | Standard | Sample |
|        |  Type   |                  |    r     |         |  Deviati |  Numbe |
|        |         |                  |          |         |    on    |   r    |
+--------+---------+------------------+----------+---------+----------+--------+
| microp | Opensmi | F0semitoneFrom27 | ['Twenti | 24.907  |  5.774   |  371   |
| hone_t | leFeatu | .5Hz_sma3nz_amea | esMale'] |         |          |        |
|  ask   |   res   |        n         |          |         |          |        |
+--------+---------+------------------+----------+---------+----------+--------+
| freesp | Opensmi | F0semitoneFrom27 | ['Twenti | 24.839  |  5.125   |  371   |
| eech_t | leFeatu | .5Hz_sma3nz_amea | esMale'] |         |          |        |
|  ask   |   res   |        n         |          |         |          |        |
+--------+---------+------------------+----------+---------+----------+--------+
| pictur | Opensmi | F0semitoneFrom27 | ['Twenti |  25.07  |   4.71   |  371   |
| e_task | leFeatu | .5Hz_sma3nz_amea | esMale'] |         |          |        |
|        |   res   |        n         |          |         |          |        |
+--------+---------+------------------+----------+---------+----------+--------+
| catego | Opensmi | F0semitoneFrom27 | ['Twenti | 25.659  |  5.474   |  371   |
| ry_tas | leFeatu | .5Hz_sma3nz_amea | esMale'] |         |          |        |
|   k    |   res   |        n         |          |         |          |        |
+--------+---------+------------------+----------+---------+----------+--------+
| letter | Opensmi | F0semitoneFrom27 | ['Twenti | 25.638  |  5.578   |  371   |
| f_task | leFeatu | .5Hz_sma3nz_amea | esMale'] |         |          |        |
|        |   res   |        n         |          |         |          |        |
+--------+---------+------------------+----------+---------+----------+--------+
| paragr | Opensmi | F0semitoneFrom27 | ['Twenti | 25.474  |  5.653   |  369   |
| aph_ta | leFeatu | .5Hz_sma3nz_amea | esMale'] |         |          |        |
|   sk   |   res   |        n         |          |         |          |        |
+--------+---------+------------------+----------+---------+----------+--------+
| ahh_ta | Opensmi | F0semitoneFrom27 | ['Twenti | 26.309  |  6.227   |  371   |
|   sk   | leFeatu | .5Hz_sma3nz_amea | esMale'] |         |          |        |
|        |   res   |        n         |          |         |          |        |
+--------+---------+------------------+----------+---------+----------+--------+
| papapa | Opensmi | F0semitoneFrom27 | ['Twenti | 25.475  |   7.77   |  371   |
| _task  | leFeatu | .5Hz_sma3nz_amea | esMale'] |         |          |        |
|        |   res   |        n         |          |         |          |        |
+--------+---------+------------------+----------+---------+----------+--------+
| pataka | Opensmi | F0semitoneFrom27 | ['Twenti | 25.626  |  7.049   |  371   |
| _task  | leFeatu | .5Hz_sma3nz_amea | esMale'] |         |          |        |
|        |   res   |        n         |          |         |          |        |
+--------+---------+------------------+----------+---------+----------+--------+
| mandog | Opensmi | F0semitoneFrom27 | ['Twenti |  24.25  |   8.59   |  371   |
| _task  | leFeatu | .5Hz_sma3nz_amea | esMale'] |         |          |        |
|        |   res   |        n         |          |         |          |        |
+--------+---------+------------------+----------+---------+----------+--------+
| tourbu | Opensmi | F0semitoneFrom27 | ['Twenti | 24.281  |  7.271   |  371   |
| s_task | leFeatu | .5Hz_sma3nz_amea | esMale'] |         |          |        |
|        |   res   |        n         |          |         |          |        |
+--------+---------+------------------+----------+---------+----------+--------+
| diagno | Opensmi | F0semitoneFrom27 | ['Twenti | 24.382  |  6.588   |  335   |
| sis_ta | leFeatu | .5Hz_sma3nz_amea | esMale'] |         |          |        |
|   sk   |   res   |        n         |          |         |          |        |
+--------+---------+------------------+----------+---------+----------+--------+
| medica | Opensmi | F0semitoneFrom27 | ['Twenti | 24.601  |  6.445   |  323   |
| tion_t | leFeatu | .5Hz_sma3nz_amea | esMale'] |         |          |        |
|  ask   |   res   |        n         |          |         |          |        |
+--------+---------+------------------+----------+---------+----------+--------+

visualizing data

Visualizing single features across tasks

Visualize the means and standard deviations for a feature with a bar chart:

python3 cli.py --command visualize --agegender TwentiesMale --vtype bar

Which returns the image below:

Visualizing cohorts and features across tasks

Visualize multiple features and references next to each other in a bar chart:

python3 cli.py --command visualize --agegender TwentiesFemale --vtype bar_cohorts --agegender ThirtiesMale

Which returns the image below:

changing settings

You can change settings.json file with the API with:

python3 cli.py --command settings 

Settings.json specifies various settings for the API. The configuration options are listed in the table below. Note all possible options are listed out in the options folder as .JSON files.

Setting Default option All Options Description
'AzureKey' 'TEST' Any string related to the Azure string Azure key for API usage in transcription.
'Features' 'OpensmileFeatures' ['OpensmileFeatures', 'ProsodyFeatures' 'PauseFeatures', 'AudiotextFeatures'] The default feature embedding to use to calculate features/references.
'FeatureType' 'F0semitoneFrom27.5Hz_sma3nz_amean' ['a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z', 'space', 'numbers', 'capletters', 'cc', 'cd', 'dt', 'ex', 'in', 'jj', 'jjr', 'jjs', 'ls', 'md', 'nn', 'nnp', 'nns', 'pdt', 'pos', 'prp', 'prp2', 'rbr', 'rbs', 'rp', 'to', 'uh', 'vb', 'vbd', 'vbg', 'vbn', 'vbp', 'vbz', 'wdt', 'wp', 'wrb', 'polarity', 'subjectivity', 'repeat', 'uniquewords', 'n_sents', 'n_words', 'n_chars', 'n_syllables', 'n_unique_words', 'n_long_words', 'n_monosyllable_words', 'n_polysyllable_words', 'flesch_kincaid_grade_level', 'flesch_reading_ease', 'smog_index', 'gunning_fog_index', 'coleman_liau_index', 'automated_readability_index', 'lix', 'gulpease_index', 'wiener_sachtextformel', 'PROPN', 'ADP', 'DET', 'NUM', 'PUNCT', 'SPACE', 'VERB', 'NOUN', 'ADV', 'CCONJ', 'PRON', 'ADJ', 'SYM', 'PART', 'INTJ', 'X', 'pos_other', 'NNP', 'IN', 'DT', 'CD', 'NNPS', ',', '_SP', 'VBZ', 'NN', 'RB', 'CC', '', 'NNS', '.', 'PRP', 'MD', 'VB', 'HYPH', 'VBD', 'JJ', ':', '-LRB-', '$', '-RRB-', 'VBG', 'VBN', 'NFP', 'RBR', 'POS', 'VBP', 'RP', 'JJS', 'PRP$', 'EX', 'JJR', 'WP', 'WDT', 'TO', 'WRB', 'PDT', 'AFX', 'RBS', 'UH', 'WP$', 'FW', 'XX', 'LS', 'ADD', 'tag_other', 'compound', 'ROOT', 'prep', 'det', 'pobj', 'nummod', 'punct', 'nsubj', 'advmod', 'conj', 'aux', 'dobj', 'nmod', 'acl', 'appos', 'npadvmod', 'amod', 'agent', 'case', 'intj', 'prt', 'pcomp', 'ccomp', 'attr', 'dep', 'acomp', 'poss', 'auxpass', 'expl', 'mark', 'nsubjpass', 'quantmod', 'advcl', 'relcl', 'oprd', 'neg', 'xcomp', 'csubj', 'predet', 'parataxis', 'dative', 'preconj', 'csubjpass', 'meta', 'dep_other', 'shape_other', 'mean sentence polarity', 'std sentence polarity', 'max sentence polarity', 'min sentence polarity', 'median sentence polarity', 'mean sentence subjectivity', 'std sentence subjectivity', 'max sentence subjectivity', 'min sentence subjectivity', 'median sentence subjectivity', 'character count', 'word count', 'sentence number', 'words per sentence', 'unique chunk noun text', 'unique chunk root text', 'unique chunk root head text', 'chunkdep ROOT', 'chunkdep pobj', 'chunkdep nsubj', 'chunkdep dobj', 'chunkdep conj', 'chunkdep appos', 'chunkdep attr', 'chunkdep nsubjpass', 'chunkdep dative', 'chunkdep pcomp', 'number of named entities', 'PERSON', 'NORP', 'FAC', 'ORG', 'GPE', 'LOC', 'PRODUCT', 'EVENT', 'WORK_OF_ART', 'LAW', 'LANGUAGE', 'DATE', 'TIME', 'PERCENT', 'MONEY', 'QUANTITY', 'ORDINAL', 'CARDINAL', 'filler ratio', 'type token ratio', 'standardized word entropy', 'question ratio', 'number ratio', 'Brunets Index', 'Honores statistic', 'datewords freq', 'word number', 'five word count', 'max word length', 'min word length', 'variance of vocabulary', 'std of vocabulary', 'sentencenum', 'periods', 'questions', 'interjections', 'repeatavg', 'filler_ratio', 'type_token_ratio', 'standardized_word_entropy', 'question_ratio', 'number_ratio', 'brunets_index', 'honores_statistic', 'pronoun_to_noun_ratio', 'F0semitoneFrom27.5Hz_sma3nz_amean', 'F0semitoneFrom27.5Hz_sma3nz_stddevNorm', 'F0semitoneFrom27.5Hz_sma3nz_percentile20.0', 'F0semitoneFrom27.5Hz_sma3nz_percentile50.0', 'F0semitoneFrom27.5Hz_sma3nz_percentile80.0', 'F0semitoneFrom27.5Hz_sma3nz_pctlrange0-2', 'F0semitoneFrom27.5Hz_sma3nz_meanRisingSlope', 'F0semitoneFrom27.5Hz_sma3nz_stddevRisingSlope', 'F0semitoneFrom27.5Hz_sma3nz_meanFallingSlope', 'F0semitoneFrom27.5Hz_sma3nz_stddevFallingSlope', 'loudness_sma3_amean', 'loudness_sma3_stddevNorm', 'loudness_sma3_percentile20.0', 'loudness_sma3_percentile50.0', 'loudness_sma3_percentile80.0', 'loudness_sma3_pctlrange0-2', 'loudness_sma3_meanRisingSlope', 'loudness_sma3_stddevRisingSlope', 'loudness_sma3_meanFallingSlope', 'loudness_sma3_stddevFallingSlope', 'jitterLocal_sma3nz_amean', 'jitterLocal_sma3nz_stddevNorm', 'shimmerLocaldB_sma3nz_amean', 'shimmerLocaldB_sma3nz_stddevNorm', 'HNRdBACF_sma3nz_amean', 'HNRdBACF_sma3nz_stddevNorm', 'logRelF0-H1-H2_sma3nz_amean', 'logRelF0-H1-H2_sma3nz_stddevNorm', 'logRelF0-H1-A3_sma3nz_amean', 'logRelF0-H1-A3_sma3nz_stddevNorm', 'F1frequency_sma3nz_amean', 'F1frequency_sma3nz_stddevNorm', 'F1bandwidth_sma3nz_amean', 'F1bandwidth_sma3nz_stddevNorm', 'F1amplitudeLogRelF0_sma3nz_amean', 'F1amplitudeLogRelF0_sma3nz_stddevNorm', 'F2frequency_sma3nz_amean', 'F2frequency_sma3nz_stddevNorm', 'F2amplitudeLogRelF0_sma3nz_amean', 'F2amplitudeLogRelF0_sma3nz_stddevNorm', 'F3frequency_sma3nz_amean', 'F3frequency_sma3nz_stddevNorm', 'F3amplitudeLogRelF0_sma3nz_amean', 'F3amplitudeLogRelF0_sma3nz_stddevNorm', 'alphaRatioV_sma3nz_amean', 'alphaRatioV_sma3nz_stddevNorm', 'hammarbergIndexV_sma3nz_amean', 'hammarbergIndexV_sma3nz_stddevNorm', 'slopeV0-500_sma3nz_amean', 'slopeV0-500_sma3nz_stddevNorm', 'slopeV500-1500_sma3nz_amean', 'slopeV500-1500_sma3nz_stddevNorm', 'alphaRatioUV_sma3nz_amean', 'hammarbergIndexUV_sma3nz_amean', 'slopeUV0-500_sma3nz_amean', 'slopeUV500-1500_sma3nz_amean', 'loudnessPeaksPerSec', 'VoicedSegmentsPerSec', 'MeanVoicedSegmentLengthSec', 'StddevVoicedSegmentLengthSec', 'MeanUnvoicedSegmentLength', 'StddevUnvoicedSegmentLength', 'Speech_Time_VADInt_1', 'Total_Time_VADInt_1', 'Pause_Time_VADInt_1', 'Pause_Percentage_VADInt_1', 'Pause_Speech_Ratio_VADInt_1', 'Mean_Pause_Length_VADInt_1', 'Pause_Variability_VADInt_1', 'Speech_Time_VADInt_2', 'Total_Time_VADInt_2', 'Pause_Time_VADInt_2', 'Pause_Percentage_VADInt_2', 'Pause_Speech_Ratio_VADInt_2', 'Mean_Pause_Length_VADInt_2', 'Pause_Variability_VADInt_2', 'Speech_Time_VADInt_3', 'Total_Time_VADInt_3', 'Pause_Time_VADInt_3', 'Pause_Percentage_VADInt_3', 'Pause_Speech_Ratio_VADInt_3', 'Mean_Pause_Length_VADInt_3', 'Pause_Variability_VADInt_3', 'UtteranceNumber', 'PauseNumber', 'AveragePauseLength', 'StdPauseLength', 'TimeToFirstPhonation', 'TimeToLastPhonation', 'UtterancePerMin', 'WordsPerMin', 'Duration'] The default feature label for sorting data.
'Task' 'microphone_task' ['microphone_task', 'freespeech_task', 'picture_task', 'category_task', 'letterf_task', 'paragraph_task', 'ahh_task', 'papapa_task', 'pataka_task', 'confrontational_task', 'nonword_task', 'recall_mandog_task', 'recall_tourbus_task', 'diagnosis_task', 'medication_task'] The default task to customize API usage.
'TranscriptEngine' "Azure" ["Azure", "DeepspeechDict", "DeepspeechNodict", "Wav2Vec"] The default transcription engine to use for audiotext_features.
'CleanAudio' True [True, False] Whether or not you should clean audio files during featurization to mono16000Hz.
'DefaultAgeGender' 'all' ['TwentiesMale', 'TwentiesFemale', 'ThirtiesMale', 'ThirtiesFemale', 'FourtiesMale]', 'FourtiesFemale', 'FiftiesMale', 'FiftiesFemale', 'SixtiesMale', 'SixtiesFemale', 'AllAgesGenders'] The default age and gender to use for reference ranges.

listening to samples

You can listen to samples (in a specific task like 'microphone_task' as specified by settings.json) with:

python3 cli.py --command samples 

Note these files are hosted on Google Drive for maximal flexibility with listening and playback.

get survey urls

You can get the survey urls with:

python3 cli.py --command urls

This is useful for cloning surveys on the SurveyLex interface.

running unit tests

You can run unit tests with:

python3 cli.py --command test 

surveys

Here are the 4 main surveys used in the Voiceome clinical study:

Survey Link Number of completions
Survey A https://app.surveylex.com/surveys/8a32cbb0-cc8a-11eb-9ea3-938cc8b6d71e, but for new studies recommend a slightly altered survey here. 6,426
Survey B https://app.surveylex.com/surveys/061da3f0-a637-11eb-bcc9-eba67643f616 1400
Survey C https://app.surveylex.com/surveys/a66494c0-a824-11ea-88c1-ab37bac1e1d4 800
Survey D https://app.surveylex.com/surveys/53737620-a637-11eb-bcc9-eba67643f616 100

Previews for each of these surveys are provided below as .GIFs so that you can quickly visualize the survey experience for clinical study participants.

Note you can clone these surveys @ https://surveylex.com and replicate our work in other patient populations.

Survey fragment Description Sample Audio
Consent form A standard consent form to opt into the Voiceome study. n/a
Microphone check A fragment and information screen to test that a microphone is able to collect voice input from a browser.
Free speech task A free speech prompt to collect extemporaneous speech from clinical study participants.
Category Naming Task A prompt that asks participants to name as many animals as they can in one minute.
Letter F Task A prompt that asks participants to name as many words that start with the letter F as they can in one minute.
Paragraph Reading Task A prompt that asks participants to read the Caterpillar passage.
Sustained Phonation Task A prompt to ask users to hold a sustained phonation of 'ahh' for as long as they can in 30 seconds.
Papapa Task A voice task to say puh-puh-puh as many times as a clinical study participant can in 10 seconds.
Pa-ta-ka Task A voice task to say puh-tah-kah as many times as a clinical study participant can in 10 seconds.
Confrontational Naming Task Name 25 images back-to-back in a session.
Nonword Naming Task Name a series of 10 nonsense words in a session.
Immediate Recall Tasks Recall 2 sentences played back in the browser immediately to test memory.
Spoken Diagnosis Task Asks individuals to name their clinical diagnoses (optional speech task).
Spoken Medication Task Asks individuals to name their current medications that they are taking (optional speech task).
Confounding Questions Survey section asking confounding questions like smoking histories, visual acuity, and other impairments. n/a
Demographic Questions A standard list of demographic questions n/a
Health Labels A standard list of health questionnaires that can be used as self-reported labels for vocal biomarker research. n/a
Fun Facts Fun facts and other fragments in the survey that can be used to increase user engagement. n/a

Cite us

If you incorporate our work in your research, feel free to reference us with the APA format below:

Schwoebel, James, et al. "A Longitudinal Normative Dataset and Protocol for Speech and Language Biomarker Research." medRxiv. 2021.

Preprint link can be found here.

Data access

We have tried to make acoustic and linguistic feature reference ranges for each task as open as possible. These are available in this repository under an Apache 2.0 license.

The sample audio and spreadsheets released in this repository are also released under an Apache 2.0 license. They were recorded by Jim Schwoebel in a controlled setting, cloning the protocol on the SurveyLex product. We did not use any sample data from participants here to protect their identities.

Also, note that OpenSMILE GeMAPS features are for research-only use. Keep this in mind if you are working on commercial deployments. Read more about their dual licensing model here. Since this repository was in the scope of research, we fall within this dual license model and can publish these references openly.

Raw audio data and .CSV labels for the Voiceome Study is accessible via a commercial license. If you are interested to license or access the Voiceome dataset, please reach out to Jim Schwoebel (VP of Data and Research @ Sonde) @ [email protected].

Collaborators

References

Voice Protocols

Health questionniares

Other references

About

πŸ₯ 🎀 The largest clinical study in the world to collect voice data labeled with health information (N>6,000 participants, 48 utterances, 80+ health labels). Preprint: https://www.medrxiv.org/content/10.1101/2021.08.16.21262125v1

Resources

License

Stars

Watchers

Forks

Packages

No packages published