Voice Activity Detection based on Mel spectrum energy. More...

Include dependency graph for mfcc_vad.h:

This graph shows which files directly or indirectly include this file:

Data Structures
struct	mfcc_vad_state
	VAD state structure. More...

Macros
#define	MFCC_VAD_NOISE_INIT_FRAMES 100
	Number of frames for fast noise floor convergence at startup (~1 s at 10 ms/frame).
#define	MFCC_VAD_NOISE_RISE_ALPHA 98
	Slow noise floor rise coefficient in Q1.15 (0.003 * 2^15).
#define	MFCC_VAD_NOISE_RISE_ALPHA_FAST 655
	Fast noise floor rise coefficient in Q1.15 (0.020 * 2^15).
#define	MFCC_VAD_ENERGY_THRESHOLD 2516582
	Energy threshold for speech detection in Q9.23 (0.30 * 2^23).
#define	MFCC_VAD_HANGOVER_FRAMES 20
	Hangover frame count to keep VAD active after last speech detection.

Functions
int	mfcc_vad_init (struct mfcc_vad_state vad, int num_mel_bins, int32_t sample_rate, struct processing_module mod)
	Initialize VAD state.
int	mfcc_vad_update (struct mfcc_vad_state vad, const int32_t mel_log)
	Process one Mel spectrum frame and update VAD decision.

Detailed Description

Voice Activity Detection based on Mel spectrum energy.

This VAD operates on the Q9.23 Mel log spectrum values produced by the MFCC component. It tracks a per-bin noise floor that follows the signal downward instantly and rises slowly, then computes a speech-weighted energy delta above the floor.

Macro Definition Documentation

◆ MFCC_VAD_ENERGY_THRESHOLD

#define MFCC_VAD_ENERGY_THRESHOLD 2516582

Energy threshold for speech detection in Q9.23 (0.30 * 2^23).

◆ MFCC_VAD_HANGOVER_FRAMES

#define MFCC_VAD_HANGOVER_FRAMES 20

Hangover frame count to keep VAD active after last speech detection.

◆ MFCC_VAD_NOISE_INIT_FRAMES

#define MFCC_VAD_NOISE_INIT_FRAMES 100

Number of frames for fast noise floor convergence at startup (~1 s at 10 ms/frame).

◆ MFCC_VAD_NOISE_RISE_ALPHA

#define MFCC_VAD_NOISE_RISE_ALPHA 98

Slow noise floor rise coefficient in Q1.15 (0.003 * 2^15).

◆ MFCC_VAD_NOISE_RISE_ALPHA_FAST

#define MFCC_VAD_NOISE_RISE_ALPHA_FAST 655

Fast noise floor rise coefficient in Q1.15 (0.020 * 2^15).

Function Documentation

◆ mfcc_vad_init()

int mfcc_vad_init	(	struct mfcc_vad_state *	vad,
		int	num_mel_bins,
		int32_t	sample_rate,
		struct processing_module *	mod )

Initialize VAD state.

Parameters

[out]	vad	Pointer to VAD state to initialize.
[in]	num_mel_bins	Number of Mel bins.
[in]	sample_rate	Audio sample rate in Hz.
[in]	mod	Processing module for memory allocation.

Returns: 0 on success, negative error code on failure.

◆ mfcc_vad_update()

int mfcc_vad_update	(	struct mfcc_vad_state *	vad,
		const int32_t *	mel_log )

Process one Mel spectrum frame and update VAD decision.