def _extract_feats(wavFiles): fbankExtractor = speakin_voice_feats.genFbankExtractor( allow_downsample=True, sample_frequency=8000, frame_length=25, frame_shift=10, high_freq=3700, low_freq=20, num_mel_bins=63) vadComputer = speakin_voice_feats.createVadComputer() for wavFile in wavFiles: fbankFeats = speakin_voice_feats.extractFeats(fbankExtractor, wavFile) vadResult = speakin_voice_feats.computeVad(vadComputer, fbankFeats) yield fbankFeats, vadResult
def _extract_feats(wavFiles): ### 提取特征 mfccExtracotr = speakin_voice_feats.genMfccExtractor(allow_downsample=True, sample_frequency=8000, frame_length=25, frame_shift=10, high_freq=3700, low_freq=20) vadComputer = speakin_voice_feats.createVadComputer() for wavFile in wavFiles: mfccFeats = speakin_voice_feats.extractFeats(mfccExtracotr, wavFile) vadresult = speakin_voice_feats.computeVad(vadComputer, mfccFeats) yield mfccFeats, vadresult
def _speakinMfcc(audioPath): rawAudio = AudioSegment.from_wav(audioPath) if rawAudio.frame_rate != 8000: raise Exception("wrong sample rate") if rawAudio.sample_width != 2: raise Exception("wrong sample width") if rawAudio.channels != 1: raise Exception("wrong channel count") mfccFeats = speakin_voice_feats.extractFeats(_mfccExtractor, audioPath) mfccFeats = speakin_voice_feats.computeDelta(_deltaExtractor, mfccFeats) mfccFeats = np.array(mfccFeats) return rawAudio, mfccFeats
def _extractFeats(wavFile): mfccFeats = speakin_voice_feats.extractFeats(_mfccExtractor, wavFile) vadResult = speakin_voice_feats.computeVad(_vadComputer, mfccFeats) return mfccFeats, vadResult
def _speakinMfcc(audioPath): mfccFeats = speakin_voice_feats.extractFeats(_mfccExtractor, audioPath) mfccFeats = speakin_voice_feats.computeDelta(_deltaExtractor, mfccFeats) mfccFeats = np.array(mfccFeats) print(mfccFeats) return mfccFeats
def _extractFeats(wavFile): fbankFeats = speakin_voice_feats.extractFeats(_fbankExtractor, wavFile) vadResult = speakin_voice_feats.computeVad(_vadComputer, fbankFeats) return fbankFeats, vadResult