chenjintao/banban
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

feat: add device diagnostics for missing voice replies

Browse Source 
Track microphone levels, short reads and PCM buffer drops. Report audio
queue admission separately from TLS writes, expose server replies and
ACKs before filtering, and capture per-round counters before timeouts.

Add diagnostic regression coverage and support POSIX hosts in WSS tests.
This commit is contained in:
chenjintao
2026-10-26 18:35:24 +08:00
parent 280b7a4149
commit db6109172c
9 changed files with 350 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ql-application/threadx/interface/custom_app/vad_det.c
+75 -5
View File
@@ -71,6 +71,55 @@ static int vad_tts_playback_finished_status = 0;
// 统计
static uint32 frame_count = 0;
static uint32 speech_frame_count = 0;
static struct {
unsigned int reads, short_reads, read_errors, zero_frames, samples, peak;
unsigned int ring_drops, ring_drop_bytes;
uint64_t abs_sum, last_log_ms, started_ms;
} mic_diag;
static void mic_diag_log(const char *stage)
{
printf("[MIC-DIAG] %s uptime_ms=%u age_ms=%u reads=%u short=%u read_errors=%u zero=%u peak=%u mean_abs=%u ring_drops=%u ring_drop_bytes=%u\n",
stage, (unsigned int)hal_os_up_time_ms(),
(unsigned int)(hal_os_up_time_ms() - mic_diag.started_ms),
mic_diag.reads, mic_diag.short_reads, mic_diag.read_errors, mic_diag.zero_frames,
mic_diag.peak, mic_diag.samples ? (unsigned int)(mic_diag.abs_sum / mic_diag.samples) : 0,
mic_diag.ring_drops, mic_diag.ring_drop_bytes);
}
/* Observe both preroll and live PCM writes; retain the caller's wakeup policy. */
static int mic_diag_write(const void *data, int len)
{
int written = hal_os_rb_write(ring_sender, data, len, 25);
if (written < len) {
mic_diag.ring_drops++;
mic_diag.ring_drop_bytes += (unsigned int)(len - (written > 0 ? written : 0));
}
return written;
}
static void mic_diag_read(const uint8_t *data, int len, int ret)
{
int i, bounded = len > VAD_FRAME_BYTES ? VAD_FRAME_BYTES : len;
unsigned int peak = 0;
mic_diag.reads++;
if (ret < 0) mic_diag.read_errors++;
if (len != VAD_FRAME_BYTES) mic_diag.short_reads++;
for (i = 0; i + 1 < bounded; i += 2) {
int sample = (int16_t)((unsigned int)data[i] | ((unsigned int)data[i + 1] << 8));
unsigned int magnitude = (unsigned int)(sample < 0 ? -sample : sample);
if (magnitude > peak) peak = magnitude;
mic_diag.abs_sum += magnitude;
mic_diag.samples++;
}
if (bounded > 0 && peak == 0) mic_diag.zero_frames++;
if (peak > mic_diag.peak) mic_diag.peak = peak;
if (mic_diag.reads == 1 || hal_os_up_time_ms() - mic_diag.last_log_ms >= 1000) {
printf("[MIC-DIAG] read ret=%d bytes=%d frame_peak=%u state=%d\n", ret, len, peak, g_state);
mic_diag_log("sampling");
mic_diag.last_log_ms = hal_os_up_time_ms();
}
}
extern int audio_packet_type;
@@ -84,7 +133,17 @@ static audio_record_handle record_hd;
static void ql_audio_record_callback(acm_audio_track_handle handle, acm_audio_record_event_t event)
{
//printf("audio_record_callback event:%d\r\n", event);
static int last_event = -1;
static uint64_t last_log_ms;
uint64_t now;
(void)handle;
if (event == AUDIO_RECORD_EVENT_TICK || event == AUDIO_RECORD_EVENT_NONE) return;
now = hal_os_up_time_ms();
if ((int)event != last_event || now - last_log_ms >= 1000) {
printf("[MIC-DIAG] callback event=%d uptime_ms=%u\n", event, (unsigned int)now);
last_event = (int)event;
last_log_ms = now;
}
}
void vad_tts_playback_finished()
@@ -206,7 +265,13 @@ int vad_audio_init(void) {
// 启动录音
int vad_audio_start(void) {
vad_audio_init();
memset(&mic_diag, 0, sizeof(mic_diag));
mic_diag.started_ms = hal_os_up_time_ms();
{
int init_ret = vad_audio_init();
printf("[MIC-DIAG] open ret=%d handle=%u uptime_ms=%u\n",
init_ret, (unsigned int)record_hd, (unsigned int)hal_os_up_time_ms());
}
frame_count = 0;
speech_frame_count = 0;
@@ -217,6 +282,7 @@ int vad_audio_start(void) {
// 停止录音
int vad_audio_stop(void) {
mic_diag_log("stop");
printf("Audio recording stopped, frames: %d, speech: %d",
frame_count, speech_frame_count);
@@ -235,6 +301,7 @@ int vad_audio_stop(void) {
// 停止录音
int vad_audio_stop_no_wait_resp(void) {
mic_diag_log("cancel");
conversation_cancel_wait();
printf("Audio recording stopped no_wait_resp, frames: %d, speech: %d",
@@ -381,7 +448,10 @@ void vad_audio_process(void) {
case STATE_RECORDING_ACTIVE:
memset(mic_buf, 0, sizeof(mic_buf));
//每次读取640字节音频数据
ql_audio_record_read(record_hd, mic_buf, &record_len);
{
int read_ret = ql_audio_record_read(record_hd, mic_buf, &record_len);
mic_diag_read(mic_buf, record_len, read_ret);
}
if (g_state == STATE_RECORDING_WAIT_SOUND) {
/* 首次开口确认前不向服务端发送静音。只在本地保留最近500ms，
@@ -392,7 +462,7 @@ void vad_audio_process(void) {
vad_preroll_count++;
}
} else {
if(hal_os_rb_write(ring_sender, (uint8*)mic_buf, record_len, 25)>0){
if(mic_diag_write(mic_buf, record_len)>0){
ql_rtos_semaphore_release(sender_sem);
}
}
@@ -437,7 +507,7 @@ void vad_audio_process(void) {
int preroll_queued = 0;
for (n = 0; n < vad_preroll_count; n++) {
size_t frame_index = (first + n) % VAD_PREROLL_FRAMES;
if (hal_os_rb_write(ring_sender, vad_preroll[frame_index], VAD_FRAME_BYTES, 25) > 0) {
if (mic_diag_write(vad_preroll[frame_index], VAD_FRAME_BYTES) > 0) {
preroll_queued = 1;
}
}