Sentence Segmentation 規劃

1. match 1,2,3,4 audio (match-based method) librosa.sequence.dtw

2. frequency domain noise reduction + slience removal (frequency-based method)

3. librosa temporal segmentation api librosa.segment.agglomerative feature clustering-based method)

4. waveform threshold (envelope) temporal-based method

5. pyAudioAnalysis->audioSegmentation.py-> silenceRemoval()