Speech Detection Model

SVM Classifier used to predict how much speech is detected in a wav file

---

Config file for the SVM

model_filename:name of file to save model to or load model from

audio_filename:path to audio file to predict

demo_mode:T/F

predict:T/F

train_model:T/F

load_model:T/F

save_model:T/F

n_mfcc:number of mfcc to compute

n_fft:number of samples in each fourier transform

hop_length:number of samples between successive frames

---

mfModel.py

Uses config file to determine what needs to be done.

When demo_mode is true it will only create a demo graph using a saved model

Otherwise the code will either create a new SVM classifer or load an old one and make a prediction

---

computeMFCC.py

Calculates the Mel Frequency Cepstral Coefficients for an audio file.

Uses the librosa library to perform the computations.