alphacep / vosk-api

Offline speech recognition API for Android, iOS, Raspberry Pi and servers with Python, Java, C# and Node
Apache License 2.0
7.7k stars 1.08k forks source link

Voice activity detection #1290

Open tim-gromeyer opened 1 year ago

tim-gromeyer commented 1 year ago

Hello, when I use a large moel 1GB< the processor load is quite high even if nothing is spoken. A solution would be a voice activity detection (short vad).
So that the VoskRecognizer only the data processed when speech was detected.

Edit: Yeah, I know that vosk_recognizer_accept_waveform return 1 if it detects silence but the cpu load is still pretty hight.

  1. Edit:

The CPU load goes up to 25% in idle mode(it processes data but nothing is spoken). If I stop the microphone input the cpu load goes down to 0%