void EmoVoiceVAD::transform (ITransformer::info info,
ssi_stream_t &stream_in,
ssi_stream_t &stream_out,
ssi_size_t xtra_stream_in_num,
ssi_stream_t xtra_stream_in[]) {
ssi_size_t sample_number = stream_in.num;
int16_t *inblock = ssi_pcast (int16_t, stream_in.ptr);
int *outblock = ssi_pcast (int, stream_out.ptr);
ssi_size_t steps = (sample_number - (EMOVOICEVAD_FRAME_SIZE - EMOVOICEVAD_FRAME_STEP)) / EMOVOICEVAD_FRAME_STEP;
if (steps <= 0) {
ssi_err ("Input vector too short (%d)", EMOVOICEVAD_FRAME_SIZE);
}
for (ssi_size_t i = 0; i < steps; i++) {
*outblock++ = dsp_vad_calc((dsp_sample_t *)voice, (dsp_vad_t *)vad, (dsp_sample_t *)(inblock + i*EMOVOICEVAD_FRAME_STEP));
}
}
int _asegment_vad(dsp_vad_t *vad, dsp_sample_t *signal, int n_samples, dsp_sample_t ***signal_segment, int **length) {
int va=0, i, out_sample;
int n_segments =0, last_va=0, samples=0;
int max_segments=MAX_SEGMENTS, segment_length = SEGMENT_LENGTH;
int delay_len=vad->sigbuf->length;
int *_length;
dsp_sample_t *voice;
dsp_sample_t **_segments;
int temp=0;
dsp_sample_t *_signal;
_signal = (dsp_sample_t *) rs_malloc(n_samples*sizeof(dsp_sample_t),"signal copy");
for (i=0;i<n_samples;i++)
_signal[i]=signal[i];
if (!vad) {
rs_warning("No voice activity detection info available!");
return -1;
}
voice = (dsp_sample_t *) rs_malloc(vad->frame_len * sizeof(dsp_sample_t),"Voice frame");
_segments = (dsp_sample_t **) rs_malloc(max_segments * sizeof(dsp_sample_t *),"Signal segments");
_length = (int *) rs_malloc(max_segments * sizeof(int),"Segment lengths");
for (i=0;i<=n_samples-VAD_FRAME_SHIFT || va >=0;i+=VAD_FRAME_SHIFT) {
if (i > n_samples-VAD_FRAME_SHIFT) {
va = dsp_vad_calc(voice,vad,NULL);
}
else {
if (i>n_samples-vad->frame_len) {
int j, new_len=i+vad->frame_len;
_signal= (dsp_sample_t *) rs_realloc(_signal,new_len*sizeof(dsp_sample_t),"Signal buffer");
for (j=n_samples;j<new_len;j++)
_signal[j]=0;
}
va = dsp_vad_calc(voice, vad, _signal+i);
}
if (va >=0 && ((va && !last_va) || (!va && last_va)) && samples >0) {
out_sample=i-(delay_len-vad->sigbuf->need_elems)*160;
if (out_sample>=n_samples)
out_sample=n_samples-1;
if (va) {
if (n_segments >= max_segments) {
max_segments += MAX_SEGMENTS;
_segments = (dsp_sample_t **) rs_realloc(_segments,max_segments * sizeof(dsp_sample_t *),"Signal segments");
_length = (int *) rs_realloc(_length,max_segments * sizeof(int),"Segment lengths");
}
_segments[n_segments] = (dsp_sample_t *) rs_malloc(segment_length * sizeof(dsp_sample_t),"Signal segment");
if (output)
fprintf(stderr,"[%d..",out_sample);
}
else {
if (output)
fprintf(stderr,"%d] ",out_sample-1);
_length[n_segments]=last_va*VAD_FRAME_SHIFT;
n_segments++;
segment_length=SEGMENT_LENGTH;
}
}
if (!va)
last_va=0;
if (va==1 && samples>0) {
if ((last_va+1)*VAD_FRAME_SHIFT > segment_length) {
segment_length += SEGMENT_LENGTH;
_segments[n_segments] = (dsp_sample_t *) rs_realloc(_segments[n_segments],segment_length * sizeof(dsp_sample_t),"Signal segment");
}
samples+=VAD_FRAME_SHIFT;
memcpy(_segments[n_segments]+last_va*VAD_FRAME_SHIFT,voice,VAD_FRAME_SHIFT*sizeof(dsp_sample_t));
last_va++;
}
temp=0;
while (va==0) {
temp++;
samples+=VAD_FRAME_SHIFT;
va = dsp_vad_calc(voice, vad, NULL);
}
}
if (last_va && samples >0) {
out_sample=i-(delay_len-vad->sigbuf->need_elems)*160;
if (out_sample>=n_samples)
out_sample=n_samples;
if (output)
fprintf(stderr,"%d] ",out_sample-1);
_length[n_segments]=last_va*VAD_FRAME_SHIFT;
n_segments++;
}
if (output)
fprintf(stderr,";\n");
rs_free(voice);
rs_free(_signal);
*signal_segment=_segments;
*length=_length;
return n_segments;
}