SpeechGenderRecognition

This is project allows to recognize from a wav file the gender of who speak.

If you have Telegram you can try the output of this project contacting: @GenderSpeechBot, this bot is open source and available here.

To install the project locally, read the follows instructions.

Requirements

• Python3

Libraries

• scikit-image
• pandas
• numpy
• matplotlib
• pyAudioAnalysis
• scipy
• sklearn
• hmmlearn
• simplejson
• eyed3
• pydub

You can install this requirements with

$ pip3 install -r requirements.txt

Usage

To run the classifiers use:

python3 main.py -r

To run the classifier with a new sample.wav use:

python3 main.py -w path/file.wav

Note: to convert any audio file into .wav file we recommend SoundConverter

To run the classifier with a new sample use:

python3 main.py -i path/file.csv

Extract new csv samples

Install R and packages

If you run Linux (debian-based) you can use the follows commands to install R and dependencies:

$ sudo apt install r-base
$ sudo apt install gfortran libsndfile1-dev libfftw3-dev
$ R

With the last command you just opened r, so you can run the follows command to install packages:

$ install.packages("tuneR", "seewave")
$ q()

Note: q() is to close the "r console".

Run the script

Fill the folder "male" and "female" in R/ with wavfile with the related gender.

Use Rscript to run the file extract_feature.r to generate a new file "my_voice.csv" (backup it before generate the new one).

Rscript extractor_feature.r

---

Credits

• Valentjno
• Helias