Created by: alanbekker

in data_utils/audio.py the method self.rms_db() computes: mean_square = np.mean(self._samples**2) return 10 * np.log10(mean_square)

and at inference time in on elf my utterances: self._samples is a vector of zeros therefore the np.log10(mean_square) obviously cause a core dump..any ideas how to treat this?

1.This makes sense receiving self._samples is a vector of zeros? 2. which epsilon do you recommend to add in order to avoid this in np.log10(mean_square+epsilon) and not harm the preprocessing stage?