Chunk time vs number of sample inconsistency.

[Rodolfo-S]

Why are the number of samples for each chunk size different from the time in milliseconds on the same chart? In the "Performance Metric" page the supported chunk sizes listed are 30, 60, and 100 milliseconds which, at 16kHz sampling rate should be 480, 960, and 1600 samples. However, the number of samples in that same column in parenthesis are listed as 512, 1024, and 1536 which correspond to 32, 64, and 96ms, respectively.

Now I'm wondering which one is incorrect. Should I be feeding the model 480 samples at a time or 512 samples?

[snakers4]

Hi @Rodolfo-S ,

Please use the sample numbers from the docstring:

silero-vad/utils_vad.py

Lines 131 to 165 in 7c671a7

	This method is used for splitting long audios into speech chunks using silero VAD
	Parameters
	----------
	audio: torch.Tensor, one dimensional
	One dimensional float torch.Tensor, other types are casted to torch if possible
	model: preloaded .jit silero VAD model
	threshold: float (default - 0.5)
	Speech threshold. Silero VAD outputs speech probabilities for each audio chunk, probabilities ABOVE this value are considered as SPEECH.
	It is better to tune this parameter for each dataset separately, but "lazy" 0.5 is pretty good for most datasets.
	sampling_rate: int (default - 16000)
	Currently silero VAD models support 8000 and 16000 sample rates
	min_speech_duration_ms: int (default - 250 milliseconds)
	Final speech chunks shorter min_speech_duration_ms are thrown out
	min_silence_duration_ms: int (default - 100 milliseconds)
	In the end of each speech chunk wait for min_silence_duration_ms before separating it
	window_size_samples: int (default - 1536 samples)
	Audio chunks of window_size_samples size are fed to the silero VAD model.
	WARNING! Silero VAD models were trained using 512, 1024, 1536 samples for 16000 sample rate and 256, 512, 768 samples for 8000 sample rate.
	Values other than these may affect model perfomance!!
	speech_pad_ms: int (default - 30 milliseconds)
	Final speech chunks are padded by speech_pad_ms each side
	return_seconds: bool (default - False)
	whether return timestamps in seconds (default - samples)
	visualize_probs: bool (default - False)
	whether draw prob hist or not

Why are the number of samples for each chunk size different from the time in milliseconds on the same chart? In the "Performance Metric" page the supported chunk sizes listed are 30, 60, and 100 milliseconds which, at 16kHz sampling rate should be 480, 960, and 1600 samples. However, the number of samples in that same column in parenthesis are listed as 512, 1024, and 1536 which correspond to 32, 64, and 96ms, respectively.

We just rounded up for simplicity of presentation.
We used to have more exact numbers of samples, but after model optimization it had to become a multiple of 2**N.