fextract_t *fextract_pitch_energy_create(int frame_len, char *m_e_params) {
fextract_t *fex;
dsp_fextract_t *mfcc;
fex = (fextract_t *) rs_malloc(sizeof(fextract_t),"emotion feature extraction data");
fex->n_features = V2_N_FEATURES;
/* Abtastrate etc. ... */
fex->samplerate = SAMPLERATE;
fex->frame_len = frame_len; /* global frame length (in frames) */
fex->pitch = pitch_create(AC_GAUSS);
fex->frame_shift = fex->frame_len - ((fex->pitch->method == AC_GAUSS? 2 : 1 ) * fex->pitch->periodsPerWindow / fex->pitch->minimumPitch - fex->pitch->dt) * fex->samplerate ; /* global shift (in frames) */
fex->hnr = NULL;
fex->vq = NULL;
/* MFCCs ... */
mfcc = (dsp_fextract_t *) rs_malloc(sizeof(dsp_fextract_t), "feature extraction data");
mfcc->type = dsp_fextype_MFCC;
mfcc->version = DSP_MK_VERSION(1, 4);
if (!dsp_mfcc_create(mfcc, m_e_params)) {
rs_free(mfcc);
mfcc = NULL;
rs_warning("Could not initialize MFCC configuration!");
}
fex->mfcc=mfcc;
return(fex);
}
EmoVoiceVAD::EmoVoiceVAD () {
int version = DSP_MK_VERSION(1,0);
vad = dsp_vad_create (version, EMOVOICEVAD_FRAME_SIZE);
voice = new dsp_sample_t[EMOVOICEVAD_FRAME_SIZE];
steps = 0;
no_va_counter = 0;
in_va_segment = false;
}
int emo_afile_segment(char *file, asegmentation_method_t *method, asegmentation_type_t type, dsp_sample_t ***signal_segment, int **segment_length) {
int size=SEGMENT_LENGTH, n_samples=0, n_segments, samples_read;
FILE *fp;
dsp_sample_t *signal=NULL;
dsp_vad_t *Vad;
signal = (dsp_sample_t *) rs_malloc(sizeof(dsp_sample_t) * size, "Signal data");
if (!method)
method = (asegmentation_method_t *) rs_malloc(sizeof(asegmentation_method_t),"Audio segmentation method");
if (strcmp(file,"-")==0)
fp = stdin;
else
fp = fopen(file,"r");
if (!fp) {
rs_warning("Cannot open file %s!",file);
return -1;
}
while ((samples_read =fread(signal+n_samples,sizeof(dsp_sample_t),BLOCKSIZE,fp)) && samples_read >0) {
n_samples+=samples_read;
if (size <= n_samples) {
size +=SEGMENT_LENGTH;
signal = (dsp_sample_t *) rs_realloc(signal,sizeof(dsp_sample_t) * size, "Signal data");
}
if (samples_read != BLOCKSIZE)
break;
}
fclose(fp);
if (type == vad && !method->vad) {
Vad = dsp_vad_create(DSP_MK_VERSION(1,0),VAD_FRAME_LEN);
method->vad = Vad;
}
n_segments = emo_asegment(method,type,signal,n_samples,signal_segment,segment_length);
if (n_segments == -1)
rs_error("Aborting during procession of file %s!",file);
rs_free(signal);
return n_segments;
}
dsp_vad_t *dsp_vad_create(int version, int frame_len)
{
dsp_vad_t *vad;
/* Parameter pruefen ... */
if (frame_len <= 0)
return(NULL);
/* ... speziellen Datenbereich erzeugen ... */
vad = rs_malloc(sizeof(dsp_vad_t), "VAD configuration data");
/* ... und gemaess Version fuellen ... */
vad->version = version;
vad->frame_len = frame_len;
vad->state = dsp_vad_no_decision;
vad->n_no_va_frames = 0;
vad->signal = rs_malloc(sizeof(dsp_sample_t) * vad->frame_len,
"VAD frame buffer");
/* optionale Eintraege - noch nicht verfuegbar! */
vad->sigbuf = NULL;
vad->ehist = NULL;
switch(vad->version) {
case DSP_MK_VERSION(1, 0):
/* ... Signalpuffer hinzufuegen ... */
vad->sigbuf = dsp_delay_create(V1_0_BUFFER_LEN,
sizeof(dsp_sample_t), vad->frame_len);
/* ... und begrenztes Energiehistogramm erzeugen */
vad->ehist = mx_histogram_create(
V1_0_EHIST_MIN, V1_0_EHIST_MAX, V1_0_EHIST_RES);
mx_histogram_limit_set(vad->ehist, V1_0_EHIST_LEN);
vad->last_idx = -1;
break;
default:
dsp_vad_destroy(vad);
return(NULL);
}
return(vad);
}
void dsp_vad_reset(dsp_vad_t *vad)
{
/* Parameter pruefen ... */
if (!vad)
return;
/* ... und gemaess Version ruecksetzen ... */
vad->state = dsp_vad_no_decision;
vad->n_no_va_frames = 0;
switch(vad->version) {
case DSP_MK_VERSION(1, 0):
dsp_delay_flush(vad->sigbuf);
mx_histogram_reset(vad->ehist);
vad->last_idx = -1;
break;
default:
return;
}
}
/**
* dsp_vad_calc(&voice[], vad, samples[])
* Fuehrt eine "voice activity detection" durch. Dazu wird in
* in Abhaengigkeit vom internen in 'vad' gepeichterten Zustand
* sowie der neuen Signaldaten 'samples[]' ein Zustandsuebergang
* durchgefuehrt und die Signaldaten im internen Puffer gespeichert.
*
* Aus diesem Puffer wird anschliessend ggf. wieder ein Signalframe
* entnommen, der i.A. um einige Frames verzoegert wurde.
*
* Liefert 1, falls ein Sprachframe aus dem Puffer extrahiert wurde,
* 0 bei einem Pausenframe und -1, falls keine Daten entnommen werden
* konnten bzw. im Fehlerfalle.
**/
int dsp_vad_calc(dsp_sample_t *voice, dsp_vad_t *vad, dsp_sample_t *samples)
{
dsp_vad_state_t mark;
/* Parameter pruefen ... */
if (!vad)
return(-1);
/* ... falls neue Daten vorliegen ... */
if (samples) {
/* ... Zustandsaenderung gemaess Version berechnen ... */
switch(vad->version) {
case DSP_MK_VERSION(1, 0):
_dsp_vad_calc_1_0(voice, vad, samples);
break;
default:
return(-1);
}
}
/* ... ggf. verzoegerten Frame zur Verarbeitung zurueckgeben ... */
if (dsp_delay_topm(NULL, &mark, vad->sigbuf) < 0)
return(-1);
/* ... Pause ... */
if (mark <= dsp_vad_silence) {
dsp_delay_pop(voice, vad->sigbuf);
return(0);
}
/* ... evtl. beginnender Sprachabschnitt ... */
else if (mark == dsp_vad_starting)
return(-1);
/* ... bzw. Sprache */
else {
dsp_delay_pop(voice, vad->sigbuf);
return(1);
}
}
/*
Segmentiert eine Liste von Dateien
Argumente:
filelist: Name der Datei, die die Dateien enthaelt, die segmentiert
werden sollen
method: dsp_vad_t* fuer voice activity detection,
int fuer frames mit fester Laenge
char* fuer Segmentierungsinfo
type: vad, fixed oder info
Rueckgabewert:
Anzahl der Signalabschnitte in signal_segment;
-1, falls ein Fehler aufgetreten ist
*/
int emo_afilelist_segment(char *filelist, asegmentation_method_t *method, asegmentation_type_t type,int audio_output){
FILE *fp;
int n_segments=0;
int *segment_length=NULL;
dsp_sample_t **signal_segment=NULL;
fp = fopen(filelist,"r");
if (!fp) {
rs_warning("Cannot open input file: %s!",filelist);
return -1;
}
switch (type) {
case vad: {
int stat=0, x=0;
char filename[STRING_LENGTH], outdir[STRING_LENGTH];
if (!method)
method = (asegmentation_method_t *) rs_malloc(sizeof(asegmentation_method_t),"Audio segmentation method");
if (!method->vad)
method->vad = dsp_vad_create(DSP_MK_VERSION(1,0),VAD_FRAME_LEN);
//rest der zeile ueberspringen
while(fscanf(fp,"%s\t%s",filename,outdir)==2 && stat >=0) {
if (output)
fprintf(stderr,"%s\t%s\t",filename,outdir);
stat=emo_afile_segment(filename,method,vad,&signal_segment,&segment_length);
if (stat == -1){
rs_warning("Segmentation of file %s failed!",filename);
return -1;
}
if (audio_output && stat != emo_afile_output(filename,outdir,stat,signal_segment,segment_length)) {
rs_warning("Output of segments of file %s failed!",filename);
return -1;
}
if (segment_length)
rs_free(segment_length);
if (signal_segment) {
int i;
for (i=0;i<stat;i++)
if (signal_segment[i])
rs_free(signal_segment[i]);
rs_free(signal_segment);
}
n_segments+=stat;
while (x != '\n' && x != EOF)
x=fgetc(fp);
x=0;
}
break;
}
case fixed: {
int stat=0, x=0;
char filename[STRING_LENGTH], outdir[STRING_LENGTH];
if (!method->segmentation_info) {
rs_warning("No segment length or shift info available!");
return -1;
}
while(fscanf(fp,"%s\t%s",filename,outdir)==2 && stat >=0) {
stat=emo_afile_segment(filename,method,fixed,&signal_segment,&segment_length);
if (stat == -1){
rs_warning("Segmentation of file %s failed!",filename);
return -1;
}
if (audio_output) {
stat = emo_afile_output(filename,outdir,stat,signal_segment,segment_length);
if (stat == -1){
rs_warning("Output of segments of file %s failed!",filename);
}
return -1;
}
n_segments+=stat;
while (x != '\n' && x != EOF)
x=fgetc(fp);
x=0;
}
break;
}
case info: {
int stat=0;
char filename[STRING_LENGTH], outdir[STRING_LENGTH], seg_info[10*STRING_LENGTH];
if (!method->segmentation_info) {
rs_warning("No segment length or shift info available!");
return -1;
}
while(fscanf(fp,"%s\t%s\t%[^\n]s",filename,outdir,seg_info)==3 && stat >=0) {
method->segmentation_info=seg_info;
stat=emo_afile_segment(filename,method,info,&signal_segment,&segment_length);
if (stat == -1){
rs_warning("Segmentation of file %s failed!",filename);
return -1;
}
if (audio_output && stat != emo_afile_output(filename,outdir,stat,signal_segment,segment_length)) {
rs_warning("Output of segments of file %s failed!",filename);
return -1;
}
n_segments+=stat;
}
break;
}
default: rs_warning("Unknown segmentation type: %d",type);
return -1;
}
fclose(fp);
return n_segments;
}
