Speaker Recognition System using CNN 🔥

A complete report of the implementation of this system is available here : https://github.com/karthxk07/SpeakerRecognition/blob/b00dbcbd3986797517dcd5773d25b91f4e5d841d/Report.pdf

For a local dataset used which contained voice samples from 2 users (with Epoch=25)

• Loss: 0.0443
• Accuracy: 0.9923
• Val_loss: 0.0622
• Val_accuracy: 0.9868

How to setup locally ?

Firstly clone the repo locally using

git clone https://github.com/karthxk07/SpeakerRecognition.git

Then run the setup script

chmod +x ./setup.sh
./setup.sh

If you have your own dataset copy it into the audio folder that is created

ensure that inside the audio folder each speaker is denoted with a different folder, which contain the speaker's voice samples.

Otherwise, you can run record.py to record 10 samples of your voice

python3 record.py

Now, Edit your train.py according to your dataset

Line 10 -

names = ["..."] //Names of all the speakers in your dataset

Also , edit predict.py accordingly

Line 12 -

names = ["..."] //Names of all the speakers in your dataset

Training and Testing

Run the train.py file now

python3 train.py

This will give the accuracy and the val_accuracy of the model

After the training is completed, now run predict.py to indentify any speaker

python3 predict.py