Personal VAD: Speaker-Conditioned Voice Activity Detection
This work addresses the need for efficient, speaker-specific voice activity detection to reduce computational cost and battery consumption in streaming on-device speech recognition, particularly where keyword detectors are not ideal.
The paper tackles the problem of detecting voice activity specifically for a target speaker at the frame level, achieving a model with only 130K parameters that outperforms a baseline combining standard VAD and speaker recognition networks.
In this paper, we propose "personal VAD", a system to detect the voice activity of a target speaker at the frame level. This system is useful for gating the inputs to a streaming on-device speech recognition system, such that it only triggers for the target user, which helps reduce the computational cost and battery consumption, especially in scenarios where a keyword detector is unpreferable. We achieve this by training a VAD-alike neural network that is conditioned on the target speaker embedding or the speaker verification score. For each frame, personal VAD outputs the probabilities for three classes: non-speech, target speaker speech, and non-target speaker speech. Under our optimal setup, we are able to train a model with only 130K parameters that outperforms a baseline system where individually trained standard VAD and speaker recognition networks are combined to perform the same task.