Neural network implemented in java that processes voice audio data and identifies specific phonemes
Originally developed as part of a EE University Project by James Richards and Dr Paul Leonard (University of Bath, UK)
There is a pre-trained neural network already in the project:
- To use a microphone as a real-time input execute the Java application at phonemeNeuralNetwork/src/speech/MainApp.java
- To read in patient voice files from /src/speech/wavFiles/patients/ update the 'fileName' variable at phonemeNeuralNetwork/src/speech/MainAppFile.java and execute as a Java application
If you want to train a new neural network:
- Load your .wav files into the directory src/speech/wavFiles/trainingSplice
- Update the training script with the new parameters
- Execute the training script at src/speech/neuralNetwork/WavTraining.java
- Wait for a while until the network is trained
- Execute the Java Application as described above