Hi, I found a mistake when run VAD on a single. If the pSpectrum value has only zeroes on first string, the result of estnoisem function return table with np.nan. I added strings
begin = 0
while begin < nFrames and np.all(pSpectrum[begin,:] == 0.0):
begin += 1
p=pSpectrum[begin,:]
to redefinition p valuation and run for cycle in range(begin, nFrames).
That improve a crashing of code, but hold that mistake
Voice_Activity_Detector/utils/estnoise_ms.py:108: RuntimeWarning: divide by zero encountered in double_scalars
acb=(1+(sum(p) / sum(pSpectrum_t)-1)**2)**(-1) # alpha_c-bar(t) (9)
Hi, I found a mistake when run VAD on a single. If the pSpectrum value has only zeroes on first string, the result of
estnoisemfunction return table withnp.nan. I added stringsto redefinition
pvaluation and runforcycle inrange(begin, nFrames).That improve a crashing of code, but hold that mistake