search

jymsuper / SpeakerRecognition_tutorial

Simple d-vector based Speaker Recognition (verification and identification) using Pytorch
MIT License
210 stars 46 forks source link

Error #16

Closed Adarsh474532 closed 3 years ago

Adarsh474532 commented 3 years ago

SIr i am unable to convert single wav to p file, can you do it please sir

jymsuper commented 3 years ago

In the function 'extract_MFB' of the below code, https://github.com/seongmin-kye/meta-SR/blob/master/feat_extract/feature_extraction.py

1) sr, audio = sio.wavfile.read(filename) => Read the audio file

2) features, energies = fbank(audio, samplerate=c.SAMPLE_RATE, nfilt=c.FILTER_BANK, winlen=0.025, winfunc=np.hamming) features = 20 * np.log10(np.maximum(features,1e-5)) => Extract features using the audio

3) features = normalize_frames(features, Scale=c.USE_SCALE) total_features = features => Normalize the features

4) speaker_label = speaker_folder # set label as a folder name (recommended). Convert this to speaker index when training feat_and_label = {'feat':total_features, 'label':speaker_label} with open(output_filename, 'wb') as fp: pickle.dump(feat_and_label, fp) => Save features and labels as a dictionary. The file name is 'output_filename' which ends in '.p' format.

Please follow the above instruction.