Extracting log mel filterbank features

[bwang482]

Thanks very much for the great library! It's my default library for speech processing now.

Just want to double check on the following, I want to extract 40-dimensional log mel filterbank feautres from sliding a Hamming window of width 25ms with an overlap of 10ms. Does the code below extract the right features? I am a bit uncertain whether frame_stride=0.01 creates overlap of 10ms..

fs, signal = wav.read(file_path)
lmfe = speechpy.feature.lmfe(signal, sampling_frequency=fs, frame_length=0.025, frame_stride=0.01, num_filters=40, fft_length=512, low_frequency=0, high_frequency=None)

Thanks!

[MichelleYang2017]

I suggest that you should read some source code:

frame_stride (float): the step between successive frames in seconds.

So,frame_stride means that no overlap