webrtcvad_voice_activity_detection_pcm_frame_processing.py

1import webrtcvad
2
3# Initialize the Voice Activity Detector
4vad = webrtcvad.Vad()
5
6# Set aggressiveness mode: 0 (least aggressive) to 3 (most aggressive)
7# Mode 3 is the most aggressive at filtering out non-speech.
8vad.set_mode(3)
9
10# Example: Process a 30ms frame of silence at 16000Hz.
11# A frame must be 10, 20, or 30 ms in duration.
12# For 16000Hz, a 30ms frame is 480 samples. 
13# Since it's 16-bit PCM (2 bytes per sample), the buffer length is 960 bytes.
14sample_rate = 16000
15frame_duration_ms = 30 # ms
16frame = b'\x00\x00' * int(sample_rate * frame_duration_ms / 1000)
17
18# Returns True if speech is detected, False otherwise
19is_speech = vad.is_speech(frame, sample_rate)
20
21print(f"Is speech detected: {is_speech}")