/**
* <EN>
* Create a new configuration instance and load parameters from a jconf
* file.
* </EN>
* <JA>
* 新たな設定インスタンスを割り付け，そこに
* jconfファイルから設定パラメータを読み込んで返す.
* </JA>
*
* @param filename [in] jconf filename
*
* @return the newly allocated global configuration instance.
*
* @callgraph
* @callergraph
* @ingroup jconf
*/
Jconf *
j_config_load_file_new(char *filename)
{
Jconf *jconf;
jconf = j_jconf_new();
if (j_config_load_file(jconf, filename) == -1) {
j_jconf_free(jconf);
return NULL;
}
return jconf;
}
/**
* <EN>
* Create a new configuration instance and load parameters from string
* file.
* </EN>
* <JA>
* 新たな設定インスタンスを割り付け，そこに
* 文字列から設定パラメータを読み込んで返す.
* </JA>
*
* @param string [in] option string
*
* @return the newly allocated global configuration instance.
*
* @callgraph
* @callergraph
* @ingroup jconf
*/
Jconf *
j_config_load_string_new(char *string)
{
Jconf *jconf;
jconf = j_jconf_new();
if (j_config_load_string(jconf, string) == -1) {
j_jconf_free(jconf);
return NULL;
}
return jconf;
}
/**
* <EN>
* Create a new configuration instance and load parameters from command
* argments.
* </EN>
* <JA>
* コマンド引数からパラメータを読み込み，その値を格納した
* 新たな設定インスタンスを割り付けて返す.
* </JA>
*
* @param argc [in] number of arguments
* @param argv [in] list of argument strings
*
* @return the newly allocated global configuration instance.
*
* @callgraph
* @callergraph
* @ingroup jconf
*/
Jconf *
j_config_load_args_new(int argc, char *argv[])
{
Jconf *jconf;
jconf = j_jconf_new();
if (j_config_load_args(jconf, argc, argv) == -1) {
j_jconf_free(jconf);
return NULL;
}
return jconf;
}
/**
* <JA>
* メイン関数
*
* @param argc [in] 引数列の長さ
* @param argv [in] 引数列
*
* @return
* </JA>エラー時 1，通常終了時 0 を返す．
* <EN>
* Main function.
*
* @param argc [in] number of argument.
* @param argv [in] array of arguments.
*
* @return 1 on error, 0 on success.
* </EN>
*/
int
main(int argc, char *argv[])
{
Recog *recog;
Jconf *jconf;
int ret;
int i;
boolean is_continues;
/* create instance */
recog = j_recog_new();
jconf = j_jconf_new();
recog->jconf = jconf;
/********************/
/* setup parameters */
/********************/
/* register additional options */
j_add_option("-in", 1, 1, "input from", opt_in);
j_add_option("-out", 1, 1, "output to", opt_out);
j_add_option("-server", 1, 1, "hostname (-out adinnet)", opt_server);
j_add_option("-NA", 1, 1, "NetAudio server host:unit (-in netaudio)", opt_NA);
j_add_option("-port", 1, 1, "port number (-out adinnet)", opt_port);
j_add_option("-inport", 1, 1, "port number (-in adinnet)", opt_inport);
j_add_option("-filename", 1, 1, "(base) filename to record (-out file)", opt_filename);
j_add_option("-startid", 1, 1, "recording start id (-out file)", opt_startid);
j_add_option("-freq", 1, 1, "sampling frequency in Hz", opt_freq);
j_add_option("-nosegment", 0, 0, "not segment input speech, record all", opt_nosegment);
j_add_option("-segment", 0, 0, "force segment input speech", opt_segment);
j_add_option("-oneshot", 0, 0, "exit after the first input", opt_oneshot);
j_add_option("-raw", 0, 0, "save in raw (BE) format", opt_raw);
j_add_option("-autopause", 0, 0, "automatically pause at each input end", opt_autopause);
j_add_option("-loosesync", 0, 0, "loose sync of resume among servers", opt_loosesync);
j_add_option("-rewind", 1, 1, "rewind to the msec", opt_rewind);
j_add_option("-h", 0, 0, "display this help", opt_help);
j_add_option("-help", 0, 0, "display this help", opt_help);
j_add_option("--help", 0, 0, "display this help", opt_help);
/* when no argument, output help and exit */
if (argc <= 1) {
opt_help(jconf, NULL, 0);
return 0;
}
/* read arguments and set parameters */
if (j_config_load_args(jconf, argc, argv) == -1) {
fprintf(stderr, "Error reading arguments\n");
return -1;
}
/* check needed arguments */
if (speech_output == SPOUT_FILE && filename == NULL) {
fprintf(stderr, "Error: output filename not specified\n");
return(-1);
}
if (speech_output == SPOUT_ADINNET && adinnet_servnum < 1) {
fprintf(stderr, "Error: adinnet server name for output not specified\n");
return(-1);
}
if (jconf->input.speech_input == SP_ADINNET &&
speech_output != SPOUT_ADINNET &&
adinnet_servnum >= 1) {
fprintf(stderr, "Warning: you specified port num by -port, but it's for output\n");
fprintf(stderr, "Warning: you may specify input port by -inport instead.\n");
fprintf(stderr, "Warning: now the default value (%d) will be used\n", ADINNET_PORT);
}
#ifdef USE_NETAUDIO
if (jconf->input.speech_input == SP_NETAUDIO && jconf->input.netaudio_devname == NULL) {
fprintf(stderr, "Error: NetAudio server name not specified\n");
return(-1);
}
#endif
if (adinnet_portnum != adinnet_servnum) {
/* if only one server, use default */
if (adinnet_servnum == 1) {
adinnet_port[0] = ADINNET_PORT;
adinnet_portnum = 1;
} else {
fprintf(stderr, "Error: you should specify both server names and different port for each!\n");
fprintf(stderr, "\tserver:");
for(i=0;i<adinnet_servnum;i++) fprintf(stderr, " %s", adinnet_serv[i]);
fprintf(stderr, "\n\tport :");
for(i=0;i<adinnet_portnum;i++) fprintf(stderr, " %d", adinnet_port[i]);
fprintf(stderr, "\n");
return(-1);
}
}
/* set Julius default parameters for unspecified acoustic parameters */
apply_para(&(jconf->am_root->analysis.para), &(jconf->am_root->analysis.para_default));
/* set some values */
jconf->input.sfreq = jconf->am_root->analysis.para.smp_freq;
jconf->input.period = jconf->am_root->analysis.para.smp_period;
jconf->input.frameshift = jconf->am_root->analysis.para.frameshift;
jconf->input.framesize = jconf->am_root->analysis.para.framesize;
/* disable successive segmentation when no segmentation available */
if (!jconf->detect.silence_cut) continuous_segment = FALSE;
/* store sampling rate locally */
sfreq = jconf->am_root->analysis.para.smp_freq;
/********************/
/* setup for output */
/********************/
if (speech_output == SPOUT_FILE) {
/* allocate work area for output file name */
if (continuous_segment) {
outpath = (char *)mymalloc(strlen(filename) + 10);
} else {
if (use_raw) {
outpath = filename;
} else {
outpath = new_output_filename(filename, ".wav");
}
}
} else if (speech_output == SPOUT_ADINNET) {
/* connect to adinnet server(s) */
for(i=0;i<adinnet_servnum;i++) {
fprintf(stderr, "connecting to #%d (%s:%d)...", i+1, adinnet_serv[i], adinnet_port[i]);
sd[i] = make_connection(adinnet_serv[i], adinnet_port[i]);
if (sd[i] < 0) return 1; /* on error */
fprintf(stderr, "connected\n");
}
} else if (speech_output == SPOUT_STDOUT) {
/* output to stdout */
fd = 1;
fprintf(stderr,"[STDOUT]");
}
/**********************/
/* interrupt handling */
/**********************/
if (signal(SIGINT, interrupt_record) == SIG_ERR) {
fprintf(stderr, "Warning: signal interruption may collapse output\n");
}
if (signal(SIGTERM, interrupt_record) == SIG_ERR) {
fprintf(stderr, "Warning: signal interruption may collapse output\n");
}
#ifdef SIGPIPE
if (signal(SIGPIPE, interrupt_record) == SIG_ERR) {
fprintf(stderr, "Warning: signal interruption may collapse output\n");
}
#endif
#ifdef SIGQUIT
if (signal(SIGQUIT, interrupt_record) == SIG_ERR) {
fprintf(stderr, "Warning: signal interruption may collapse output\n");
}
#endif
/***************************/
/* initialize input device */
/***************************/
if (jconf->input.speech_input == SP_ADINNET) {
jconf->input.adinnet_port = adinnet_port_in;
}
if (j_adin_init(recog) == FALSE) {
fprintf(stderr, "Error in initializing adin device\n");
return -1;
}
if (rewind_msec > 0) {
/* allow adin module to keep triggered speech while pausing */
#ifdef HAVE_PTHREAD
if (recog->adin->enable_thread) {
recog->adin->ignore_speech_while_recog = FALSE;
}
#endif
}
/*********************/
/* add some callback */
/*********************/
callback_add(recog, CALLBACK_EVENT_SPEECH_START, record_trigger_time, NULL);
/**************************************/
/* display input/output configuration */
/**************************************/
put_status(recog);
/*******************/
/* begin recording */
/*******************/
if (continuous_segment) { /* reset parameter for successive output */
total_speechlen = 0;
sid = startid;
}
fprintf(stderr,"[start recording]\n");
if (jconf->input.speech_input == SP_RAWFILE) file_counter = 0;
/*********************/
/* input stream loop */
/*********************/
while(1) {
/* begin A/D input of a stream */
ret = j_open_stream(recog, NULL);
switch(ret) {
case 0: /* succeeded */
break;
case -1: /* error */
/* go on to next input */
continue;
case -2: /* end of recognition process */
switch(jconf->input.speech_input) {
case SP_RAWFILE:
fprintf(stderr, "%d files processed\n", file_counter);
break;
case SP_STDIN:
fprintf(stderr, "reached end of input on stdin\n");
break;
default:
fprintf(stderr, "failed to begin input stream\n");
}
/* exit recording */
goto record_end;
}
/*********************************/
/* do VAD and recording */
/*********************************/
do {
/* process one segment with segmentation */
/* for incoming speech input, speech detection and segmentation are
performed and, adin_callback_* is called for speech output for each segment block.
*/
/* adin_go() return when input segmented by long silence, or input stream reached to the end */
speechlen = 0;
stop_at_next = FALSE;
if (jconf->input.speech_input == SP_MIC) {
fprintf(stderr, "<<< please speak >>>");
}
if (speech_output == SPOUT_ADINNET) {
ret = adin_go(adin_callback_adinnet, adinnet_check_command, recog);
} else {
ret = adin_go(adin_callback_file, NULL, recog);
}
/* return value of adin_go:
-2: input terminated by pause command from adinnet server
-1: input device read error or callback process error
0: paused by input stream (end of file, etc..)
>0: detected end of speech segment:
by adin-cut, or by callback process
(or return value of ad_check (<0) (== not used in this program))
*/
/* if PAUSE or TERMINATE command has been received while input,
stop_at_next is TRUE here */
switch(ret) {
case -2: /* terminated by terminate command from server */
fprintf(stderr, "[terminated by server]\n");
break;
case -1: /* device read error or callback error */
fprintf(stderr, "[error]\n");
break;
case 0: /* reached to end of input */
fprintf(stderr, "[eof]\n");
break;
default: /* input segmented by silence or callback process */
fprintf(stderr, "[segmented]\n");
break;
}
if (ret == -1) {
/* error in input device or callback function, so terminate program here */
return 1;
}
/*************************/
/* one segment processed */
/*************************/
if (speech_output == SPOUT_FILE) {
/* close output files */
if (close_files() == FALSE) return 1;
} else if (speech_output == SPOUT_ADINNET) {
if (speechlen > 0) {
if (ret >= 0 || stop_at_next) { /* segmented by adin-cut or end of stream or server-side command */
/* send end-of-segment ack to client */
adin_send_end_of_segment();
}
/* output info */
printf("sent: %d samples (%.2f sec.) [%6d (%5.2fs) - %6d (%5.2fs)]\n",
speechlen, (float)speechlen / (float)sfreq,
trigger_sample, (float)trigger_sample / (float)sfreq,
trigger_sample + speechlen, (float)(trigger_sample + speechlen) / (float)sfreq);
}
}
/*************************************/
/* increment ID and total sample len */
/*************************************/
if (continuous_segment) {
total_speechlen += speechlen;
sid++;
}
/***************************************************/
/* with adinnet server, if terminated by */
/* server-side PAUSE command, wait for RESUME here */
/***************************************************/
if (pause_each) {
/* pause at each end */
//if (speech_output == SPOUT_ADINNET && speechlen > 0) {
if (speech_output == SPOUT_ADINNET) {
if (adinnet_wait_command() < 0) {
/* command error: terminate program here */
return 1;
}
}
} else {
if (speech_output == SPOUT_ADINNET && stop_at_next) {
if (adinnet_wait_command() < 0) {
/* command error: terminate program here */
return 1;
}
}
}
/* loop condition check */
is_continues = FALSE;
if (continuous_segment && (ret > 0 || ret == -2)) {
is_continues = TRUE;
}
} while (is_continues); /* to the next segment in this input stream */
/***********************/
/* end of input stream */
/***********************/
adin_end(recog->adin);
} /* to the next input stream (i.e. next input file in SP_RAWFILE) */
record_end:
if (speech_output == SPOUT_FILE) {
if (continuous_segment) {
printf("recorded total %d samples (%.2f sec.) segmented to %s.%04d - %s.%04d files\n", total_speechlen, (float)total_speechlen / (float)sfreq, filename, 0, filename, sid-1);
}
}
return 0;
}
int
main(int argc, char *argv[])
{
Recog *recog;
Jconf *jconf;
float *ss;
MFCCWork *wrk;
/* create instance */
recog = j_recog_new();
jconf = j_jconf_new();
recog->jconf = jconf;
/* set application-specific additional options */
j_add_option("-freq", 1, 1, "sampling freq in Hz", opt_freq);
j_add_option("-len", 1, 1, "record length in msec", opt_len);
j_add_option("-h", 0, 0, "display this help", opt_help);
j_add_option("-help", 0, 0, "display this help", opt_help);
j_add_option("--help", 0, 0, "display this help", opt_help);
/* when no argument, output help and exit */
if (argc <= 1) {
opt_help(jconf, NULL, 0);
return 0;
}
/* regard last arg as filename */
if (strmatch(argv[argc-1], "-")) {
stout = TRUE;
} else {
filename = argv[argc-1];
}
/* set default as same as "-input mic" */
jconf->input.type = INPUT_WAVEFORM;
jconf->input.speech_input = SP_MIC;
jconf->input.device = SP_INPUT_DEFAULT;
/* process config and load them */
if (j_config_load_args(jconf, argc-1, argv) == -1) {
fprintf(stderr, "Error reading arguments\n");
return -1;
}
/* force some default values */
jconf->detect.silence_cut = 0; /* disable silence cut */
jconf->preprocess.strip_zero_sample = TRUE; /* strip zero samples */
jconf->detect.level_thres = 0; /* no VAD, record all */
/* set Julius default parameters for unspecified acoustic parameters */
apply_para(&(jconf->am_root->analysis.para), &(jconf->am_root->analysis.para_default));
/* set some values */
jconf->input.sfreq = jconf->am_root->analysis.para.smp_freq;
jconf->input.period = jconf->am_root->analysis.para.smp_period;
jconf->input.frameshift = jconf->am_root->analysis.para.frameshift;
jconf->input.framesize = jconf->am_root->analysis.para.framesize;
sfreq = jconf->am_root->analysis.para.smp_freq;
/* output file check */
if (!stout) {
if (access(filename, F_OK) == 0) {
if (access(filename, W_OK) == 0) {
fprintf(stderr, "Warning: overwriting file \"%s\"\n", filename);
} else {
perror("mkss");
return(1);
}
}
}
/* allocate speech store buffer */
samples = sfreq * slen / 1000;
speech = (SP16 *)mymalloc(sizeof(SP16) * samples);
/* allocate work area to compute spectrum */
wrk = WMP_work_new(&(jconf->am_root->analysis.para));
if (wrk == NULL) {
jlog("ERROR: m_fusion: failed to initialize MFCC computation for SS\n");
return -1;
}
/* initialize input device */
if (j_adin_init(recog) == FALSE) {
fprintf(stderr, "Error in initializing adin device\n");
return -1;
}
/* open device */
if (j_open_stream(recog, NULL) < 0) {
fprintf(stderr, "Error in opening adin device\n");
}
/* record mic input */
fprintf(stderr, "%dHz recording for %.2f seconds of noise\n", sfreq, (float)slen /(float)1000);
speechnum = 0;
adin_go(adin_callback, NULL, recog);
/* close device */
adin_end(recog->adin);
fprintf(stderr, "\n%d samples (%d bytes, %.1f sec) recorded\n", samples, samples * sizeof(SP16), (float)samples / (float)sfreq);
/* compute SS */
fprintf(stderr, "compute SS:\n");
fprintf(stderr, " fsize : %4d samples (%.1f msec)\n", jconf->input.framesize, (float)jconf->input.framesize * 1000.0/ (float)sfreq);
fprintf(stderr, " fshift: %4d samples (%.1f msec)\n", jconf->input.frameshift, (float)jconf->input.frameshift * 1000.0/ (float)sfreq);
ss = new_SS_calculate(speech, samples, &sslen, wrk, &(jconf->am_root->analysis.para));
fprintf(stderr, " points: %4d\n", sslen);
fprintf(stderr, "noise spectrum was measured\n");
/* open file for recording */
fprintf(stderr, "writing average noise spectrum to [%s]...", filename);
if (stout) {
fd = 1;
} else {
if ((fd = open(filename, O_CREAT | O_RDWR
#ifdef O_BINARY
| O_BINARY
#endif
, 0644)) == -1) {
perror("mkss");
return(1);
}
}
x = sslen;
#ifndef WORDS_BIGENDIAN
swap_bytes((char *)&x, sizeof(int), 1);
#endif
if (write(fd, &x, sizeof(int)) < sizeof(int)) {
perror("mkss");
return(1);
}
#ifndef WORDS_BIGENDIAN
swap_bytes((char *)ss, sizeof(float), sslen);
#endif
if (write(fd, ss, sslen * sizeof(float)) < sslen * sizeof(float)) {
perror("mkss");
return(1);
}
if (!stout) {
if (close(fd) < 0) {
perror("mkss");
return(1);
}
}
fprintf(stderr, "done\n");
WMP_free(wrk);
return 0;
}
int cTumkwsjSink::setupJulius()
{
try {
int argc=3;
char* argv[3] = {"julius","-C",NULL};
if (configfile != NULL)
argv[2]=strdup(configfile);
else
argv[2]=strdup("kws.cfg");
/* add application option dummies */
/*
j_add_option("-gprob", 1, 1, "garbage probability", opt_gprob);
j_add_option("-kprob", 1, 1, "keyword probability", opt_kprob);
j_add_option("-knum", 1, 1, "number of keywords", opt_knum);
*/
/* create a configuration variables container */
jconf = j_jconf_new();
// j_config_load_file(jconf, strdup(configfile));
if (j_config_load_args(jconf, argc, argv) == -1) {
COMP_ERR("error parsing julius decoder options, this might be a bug. see tumkwsjSink.cpp!");
j_jconf_free(jconf);
free(argv[2]);
return 0;
}
free(argv[2]);
/* output system log to a file */
if (getInt("debug") != 1) {
jlog_set_output(NULL);
}
/* here you can set/modify any parameter in the jconf before setup */
jconf->input.type = INPUT_VECTOR;
jconf->input.speech_input = SP_EXTERN;
jconf->decodeopt.realtime_flag = TRUE; // ???
jconf->ext.period = (float)(reader->getLevelT());
jconf->ext.veclen = reader->getLevelN();
jconf->ext.userptr = (void *)this;
jconf->ext.fv_read = external_fv_read_loader;
/* Fixate jconf parameters: it checks whether the jconf parameters
are suitable for recognition or not, and set some internal
parameters according to the values for recognition. Modifying
a value in jconf after this function may be errorous.
*/
if (j_jconf_finalize(jconf) == FALSE) {
SMILE_IERR(1,"error finalising julius jconf in setupJulius()!");
j_jconf_free(jconf);
return 0;
}
/* create a recognition instance */
recog = j_recog_new();
/* use user-definable data hook to set a pointer to our class instance */
recog->hook = (void *)this;
/* assign configuration to the instance */
recog->jconf = jconf;
/* load all files according to the configurations */
if (j_load_all(recog, jconf) == FALSE) {
SMILE_IERR(1, "Error in loading model for Julius decoder");
j_recog_free(recog);
return 0;
}
SMILE_IMSG(2,"garbage prob.: %f",glogprob);
SMILE_IMSG(2,"keyword prob.: %f",klogprob);
SMILE_IMSG(2,"number of phonemes: %i",numphon);
// register user LM, get vocab size, and init lmWeights with zero or load from file
PROCESS_LM *lm;
for(lm=recog->lmlist;lm;lm=lm->next) {
if (lm->lmtype == LM_PROB) {
lm->winfo->userptr = (void*)this; // PATCH: sent/vocabulary.h (WORD_INFO struct): ++ void * userptr; // Pointer to userdata....
j_regist_user_lm_func(lm, userlm_uni_loader, userlm_bi_loader, userlm_lm_loader);
if ((numWords==0)&&((long)(lm->winfo->num)>0)) {
lmWinfo = lm->winfo;
numWords = (long)(lm->winfo->num);
}
}
}
// load lmWeights data:
//printf("XXX HEREA");
if (dynamicLm) {
//printf("XXX HERE0");
if (lmWinfo != NULL) {
//printf("XXX HERE");
lmWeights = (LOGPROB*)malloc(sizeof(LOGPROB)*numWords);
if (lmWeights == NULL) { OUT_OF_MEMORY; }
int i;
for (i=0; i<numWords; i++) { lmWeights[i] = (LOGPROB)lmpenalty; }
const char *lmfile = getStr("lmfile");
if (lmfile != NULL) {
FILE *lf=fopen(lmfile,"r");
if (lf == NULL) { SMILE_IERR(1,"Error opening word weights file (lmfile) '%s'",lmfile); }
else {
SMILE_IMSG(1,"lmfile: '%s'",lmfile);
long lineNr = 0;
char *line=NULL; size_t len=0;
size_t r=-1;
do {
r = getline(&line, &len, lf);
//printf("XXX LINE '%s'",line);
if ((r != (size_t)-1)&&(line!=NULL)) {
lineNr++;
parseLmWeightsLine(line,lineNr,lmfile);
}
} while (r != (size_t)-1);
}
}
} else {
SMILE_IERR(1,"no language model word info (vocabulary) found, check julius config!");
}
}
//----
/* checkout for recognition: build lexicon tree, allocate cache */
if (j_final_fusion(recog) == FALSE) {
fprintf(stderr, "ERROR: Error while setup work area for recognition\n");
j_recog_free(recog);
if (logfile) fclose(fp);
return 0;
}
setupCallbacks(recog, NULL);
/* output system information to log */
j_recog_info(recog);
terminated = FALSE;
}
catch (int) { }
juliusIsSetup=1;
return 1;
}
/**
* <JA>
* ヘルプを表示する.
*
* </JA>
* <EN>
* Output help document.
*
* </EN>
*
* @param fp [in] file pointer to output help
*
* @callgraph
* @callergraph
* @ingroup engine
*
*/
void
j_output_argument_help(FILE *fp)
{
Jconf *jconf;
#ifdef ENABLE_PLUGIN
int id;
char buf[64];
PLUGIN_ENTRY *p;
FUNC_VOID func;
#endif
/* load default values */
jconf = j_jconf_new();
j_put_header(fp);
j_put_compile_defs(fp);
fprintf(fp, "\nOptions:\n");
fprintf(fp, "\n--- Global Options -----------------------------------------------\n");
fprintf(fp, "\n Speech Input:\n");
fprintf(fp, " (Can extract only MFCC based features from waveform)\n");
fprintf(fp, " [-input devname] input source (default = htkparam)\n");
fprintf(fp, " htkparam/mfcfile HTK parameter file\n");
fprintf(fp, " file/rawfile waveform file (%s)\n", SUPPORTED_WAVEFILE_FORMAT);
#ifdef USE_MIC
fprintf(fp, " mic default microphone device\n");
# ifdef HAS_ALSA
fprintf(fp, " alsa use ALSA interface\n");
# endif
# ifdef HAS_OSS
fprintf(fp, " oss use OSS interface\n");
# endif
# ifdef HAS_ESD
fprintf(fp, " esd use ESounD interface\n");
# endif
#endif
#ifdef USE_NETAUDIO
fprintf(fp, " netaudio DatLink/NetAudio server\n");
#endif
fprintf(fp, " adinnet adinnet client (TCP/IP)\n");
fprintf(fp, " stdin standard input\n");
#ifdef ENABLE_PLUGIN
if (global_plugin_list) {
if ((id = plugin_get_id("adin_get_optname")) >= 0) {
for(p=global_plugin_list[id];p;p=p->next) {
func = (FUNC_VOID) p->func;
(*func)(buf, (int)64);
fprintf(fp, " %-18s(adin plugin #%d)\n", buf, p->source_id);
}
}
if ((id = plugin_get_id("fvin_get_optname")) >= 0) {
for(p=global_plugin_list[id];p;p=p->next) {
func = (FUNC_VOID) p->func;
(*func)(buf, (int)64);
fprintf(fp, " %-18s(feature vector input plugin #%d)\n", buf, p->source_id);
}
}
}
#endif
fprintf(fp, " [-filelist file] filename of input file list\n");
#ifdef USE_NETAUDIO
fprintf(fp, " [-NA host:unit] get audio from NetAudio server at host:unit\n");
#endif
fprintf(fp, " [-adport portnum] adinnet port number to listen (%d)\n", jconf->input.adinnet_port);
fprintf(fp, " [-48] enable 48kHz sampling with internal down sampler (OFF)\n");
fprintf(fp, " [-zmean/-nozmean] enable/disable DC offset removal (OFF)\n");
fprintf(fp, " [-nostrip] disable stripping off zero samples\n");
fprintf(fp, " [-record dir] record triggered speech data to dir\n");
fprintf(fp, " [-rejectshort msec] reject an input shorter than specified\n");
#ifdef POWER_REJECT
fprintf(fp, " [-powerthres value] rejection threshold of average power (%.1f)\n", jconf->reject.powerthres);
#endif
fprintf(fp, "\n Speech Detection: (default: on=mic/net off=files)\n");
/*fprintf(fp, " [-pausesegment] turn on (force) pause detection\n");*/
/*fprintf(fp, " [-nopausesegment] turn off (force) pause detection\n");*/
fprintf(fp, " [-cutsilence] turn on (force) skipping long silence\n");
fprintf(fp, " [-nocutsilence] turn off (force) skipping long silence\n");
fprintf(fp, " [-lv unsignedshort] input level threshold (0-32767) (%d)\n", jconf->detect.level_thres);
fprintf(fp, " [-zc zerocrossnum] zerocross num threshold per sec. (%d)\n", jconf->detect.zero_cross_num);
fprintf(fp, " [-headmargin msec] header margin length in msec. (%d)\n", jconf->detect.head_margin_msec);
fprintf(fp, " [-tailmargin msec] tail margin length in msec. (%d)\n", jconf->detect.tail_margin_msec);
fprintf(fp, "\n GMM utterance verification:\n");
fprintf(fp, " -gmm filename GMM definition file\n");
fprintf(fp, " -gmmnum num GMM Gaussian pruning num (%d)\n", jconf->reject.gmm_gprune_num);
fprintf(fp, " -gmmreject string comma-separated list of noise model name to reject\n");
#ifdef GMM_VAD
fprintf(fp, "\n GMM-based VAD:\n");
fprintf(fp, " -gmmmargin frames backstep margin on speech trigger (%d)\n", jconf->detect.gmm_margin);
fprintf(fp, " -gmmup score up-trigger threshold (%.1f)\n", jconf->detect.gmm_uptrigger_thres);
fprintf(fp, " -gmmdown score down-trigger threshold (%.1f)\n", jconf->detect.gmm_downtrigger_thres);
#endif
fprintf(fp, "\n On-the-fly Decoding: (default: on=mic/net off=files)\n");
fprintf(fp, " [-realtime] turn on, input streamed with MAP-CMN\n");
fprintf(fp, " [-norealtime] turn off, input buffered with sentence CMN\n");
fprintf(fp, "\n Others:\n");
fprintf(fp, " [-C jconffile] load options from jconf file\n");
fprintf(fp, " [-quiet] reduce output to only word string\n");
fprintf(fp, " [-demo] equal to \"-quiet -progout\"\n");
fprintf(fp, " [-debug] (for debug) dump numerous log\n");
fprintf(fp, " [-callbackdebug] (for debug) output message per callback\n");
fprintf(fp, " [-check (wchmm|trellis)] (for debug) check internal structure\n");
fprintf(fp, " [-check triphone] triphone mapping check\n");
fprintf(fp, " [-setting] print engine configuration and exit\n");
fprintf(fp, " [-help] print this message and exit\n");
fprintf(fp, "\n--- Instance Declarations ----------------------------------------\n\n");
fprintf(fp, " [-AM] start a new acoustic model instance\n");
fprintf(fp, " [-LM] start a new language model instance\n");
fprintf(fp, " [-SR] start a new recognizer (search) instance\n");
fprintf(fp, " [-AM_GMM] start an AM feature instance for GMM\n");
fprintf(fp, " [-GLOBAL] start a global section\n");
fprintf(fp, " [-nosectioncheck] disable option location check\n");
fprintf(fp, "\n--- Acoustic Model Options (-AM) ---------------------------------\n");
fprintf(fp, "\n Acoustic analysis:\n");
fprintf(fp, " [-htkconf file] load parameters from the HTK Config file\n");
fprintf(fp, " [-smpFreq freq] sample period (Hz) (%ld)\n", jconf->am_root->analysis.para_default.smp_freq);
fprintf(fp, " [-smpPeriod period] sample period (100ns) (%ld)\n", jconf->am_root->analysis.para_default.smp_period);
fprintf(fp, " [-fsize sample] window size (sample) (%d)\n", jconf->am_root->analysis.para_default.framesize);
fprintf(fp, " [-fshift sample] frame shift (sample) (%d)\n", jconf->am_root->analysis.para_default.frameshift);
fprintf(fp, " [-preemph] pre-emphasis coef. (%.2f)\n", jconf->am_root->analysis.para_default.preEmph);
fprintf(fp, " [-fbank] number of filterbank channels (%d)\n", jconf->am_root->analysis.para_default.fbank_num);
fprintf(fp, " [-ceplif] cepstral liftering coef. (%d)\n", jconf->am_root->analysis.para_default.lifter);
fprintf(fp, " [-rawe] [-norawe] toggle using raw energy (no)\n");
fprintf(fp, " [-enormal] [-noenormal] toggle normalizing log energy (no)\n");
fprintf(fp, " [-escale] scaling log energy for enormal (%.1f)\n", jconf->am_root->analysis.para_default.escale);
fprintf(fp, " [-silfloor] energy silence floor in dB (%.1f)\n", jconf->am_root->analysis.para_default.silFloor);
fprintf(fp, " [-delwin frame] delta windows length (frame) (%d)\n", jconf->am_root->analysis.para_default.delWin);
fprintf(fp, " [-accwin frame] accel windows length (frame) (%d)\n", jconf->am_root->analysis.para_default.accWin);
fprintf(fp, " [-hifreq freq] freq. of upper band limit, off if <0 (%d)\n", jconf->am_root->analysis.para_default.hipass);
fprintf(fp, " [-lofreq freq] freq. of lower band limit, off if <0 (%d)\n", jconf->am_root->analysis.para_default.lopass);
fprintf(fp, " [-sscalc] do spectral subtraction (file input only)\n");
fprintf(fp, " [-sscalclen msec] length of head silence for SS (msec) (%d)\n", jconf->am_root->frontend.sscalc_len);
fprintf(fp, " [-ssload filename] load constant noise spectrum from file for SS\n");
fprintf(fp, " [-ssalpha value] alpha coef. for SS (%f)\n", jconf->am_root->frontend.ss_alpha);
fprintf(fp, " [-ssfloor value] spectral floor for SS (%f)\n", jconf->am_root->frontend.ss_floor);
fprintf(fp, " [-zmeanframe/-nozmeanframe] frame-wise DC removal like HTK(OFF)\n");
fprintf(fp, " [-usepower/-nousepower] use power in fbank analysis (OFF)\n");
fprintf(fp, " [-cmnload file] load initial CMN param from file on startup\n");
fprintf(fp, " [-cmnsave file] save CMN param to file after each input\n");
fprintf(fp, " [-cmnnoupdate] not update CMN param while recog. (use with -cmnload)\n");
fprintf(fp, " [-cmnmapweight] weight value of initial cm for MAP-CMN (%6.2f)\n", jconf->am_root->analysis.cmn_map_weight);
fprintf(fp, " [-cvn] cepstral variance normalisation (%s)\n", jconf->amnow->analysis.para.cvn ? "on" : "off");
fprintf(fp, " [-vtln alpha lowcut hicut] enable VTLN (1.0 to disable) (%f)\n", jconf->am_root->analysis.para_default.vtln_alpha);
fprintf(fp, "\n Acoustic Model:\n");
fprintf(fp, " -h hmmdefsfile HMM definition file name\n");
fprintf(fp, " [-hlist HMMlistfile] HMMlist filename (must for triphone model)\n");
fprintf(fp, " [-iwcd1 methodname] switch IWCD triphone handling on 1st pass\n");
fprintf(fp, " best N use N best score (default of n-gram, N=%d)\n", jconf->am_root->iwcdmaxn);
fprintf(fp, " max use maximum score\n");
fprintf(fp, " avg use average score (default of dfa)\n");
fprintf(fp, " [-force_ccd] force to handle IWCD\n");
fprintf(fp, " [-no_ccd] don't handle IWCD\n");
fprintf(fp, " [-notypecheck] don't check input parameter type\n");
fprintf(fp, " [-spmodel HMMname] name of short pause model (\"%s\")\n", SPMODEL_NAME_DEFAULT);
fprintf(fp, " [-multipath] switch decoding for multi-path HMM (auto)\n");
fprintf(fp, "\n Acoustic Model Computation Method:\n");
fprintf(fp, " [-gprune methodname] select Gaussian pruning method:\n");
#ifdef GPRUNE_DEFAULT_SAFE
fprintf(fp, " safe safe pruning (default for TM/PTM)\n");
#else
fprintf(fp, " safe safe pruning\n");
#endif
#if GPRUNE_DEFAULT_HEURISTIC
fprintf(fp, " heuristic heuristic pruning (default for TM/PTM)\n");
#else
fprintf(fp, " heuristic heuristic pruning\n");
#endif
#if GPRUNE_DEFAULT_BEAM
fprintf(fp, " beam beam pruning (default for TM/PTM)\n");
#else
fprintf(fp, " beam beam pruning\n");
#endif
fprintf(fp, " none no pruning (default for non tmix models)\n");
#ifdef ENABLE_PLUGIN
if (global_plugin_list) {
if ((id = plugin_get_id("calcmix_get_optname")) >= 0) {
for(p=global_plugin_list[id];p;p=p->next) {
func = (FUNC_VOID) p->func;
(*func)(buf, (int)64);
fprintf(fp, " %-14s(calculation plugin #%d)\n", buf, p->source_id);
}
}
}
#endif
fprintf(fp, " [-tmix gaussnum] Gaussian num threshold per mixture for pruning (%d)\n", jconf->am_root->mixnum_thres);
fprintf(fp, " [-gshmm hmmdefs] monophone hmmdefs for GS\n");
fprintf(fp, " [-gsnum N] N-best state will be selected (%d)\n", jconf->am_root->gs_statenum);
fprintf(fp, "\n--- Language Model Options (-LM) ---------------------------------\n");
fprintf(fp, "\n N-gram:\n");
fprintf(fp, " -d file.bingram n-gram file in Julius binary format\n");
fprintf(fp, " -nlr file.arpa forward n-gram file in ARPA format\n");
fprintf(fp, " -nrl file.arpa backward n-gram file in ARPA format\n");
fprintf(fp, " [-lmp float float] weight and penalty (tri: %.1f %.1f mono: %.1f %1.f)\n", DEFAULT_LM_WEIGHT_TRI_PASS1, DEFAULT_LM_PENALTY_TRI_PASS1, DEFAULT_LM_WEIGHT_MONO_PASS1, DEFAULT_LM_PENALTY_MONO_PASS1);
fprintf(fp, " [-lmp2 float float] for 2nd pass (tri: %.1f %.1f mono: %.1f %1.f)\n", DEFAULT_LM_WEIGHT_TRI_PASS2, DEFAULT_LM_PENALTY_TRI_PASS2, DEFAULT_LM_WEIGHT_MONO_PASS2, DEFAULT_LM_PENALTY_MONO_PASS2);
fprintf(fp, " [-transp float] penalty for transparent word (%+2.1f)\n", jconf->search_root->lmp.lm_penalty_trans);
fprintf(fp, "\n DFA Grammar:\n");
fprintf(fp, " -dfa file.dfa DFA grammar file\n");
fprintf(fp, " -gram file[,file2...] (list of) grammar prefix(es)\n");
fprintf(fp, " -gramlist filename filename of grammar list\n");
fprintf(fp, " [-penalty1 float] word insertion penalty (1st pass) (%.1f)\n", jconf->search_root->lmp.penalty1);
fprintf(fp, " [-penalty2 float] word insertion penalty (2nd pass) (%.1f)\n", jconf->search_root->lmp.penalty2);
fprintf(fp, "\n Word Dictionary for N-gram and DFA:\n");
fprintf(fp, " -v dictfile dictionary file name\n");
fprintf(fp, " [-silhead wordname] (n-gram) beginning-of-sentence word (%s)\n", BEGIN_WORD_DEFAULT);
fprintf(fp, " [-siltail wordname] (n-gram) end-of-sentence word (%s)\n", END_WORD_DEFAULT);
fprintf(fp, " [-mapunk wordname] (n-gram) map unknown words to this (%s)\n", UNK_WORD_DEFAULT);
fprintf(fp, " [-forcedict] ignore error entry and keep running\n");
fprintf(fp, " [-iwspword] (n-gram) add short-pause word for inter-word CD sp\n");
fprintf(fp, " [-iwspentry entry] (n-gram) word entry for \"-iwspword\" (%s)\n", IWSPENTRY_DEFAULT);
fprintf(fp, "\n Isolated Word Recognition:\n");
fprintf(fp, " -w file[,file2...] (list of) wordlist file name(s)\n");
fprintf(fp, " -wlist filename file that contains list of wordlists\n");
fprintf(fp, " -wsil head tail sp name of silence/pause model\n");
fprintf(fp, " head - BOS silence model name (%s)\n", jconf->lm_root->wordrecog_head_silence_model_name);
fprintf(fp, " tail - EOS silence model name (%s)\n", jconf->lm_root->wordrecog_tail_silence_model_name);
fprintf(fp, " sp - their name as context or \"NULL\" (%s)\n", (jconf->lm_root->wordrecog_silence_context_name[0] == '\0') ? "NULL" : jconf->lm_root->wordrecog_silence_context_name);
#ifdef DETERMINE
fprintf(fp, " -wed float int thresholds for early word determination\n");
fprintf(fp, " float: score threshold (%.1f)\n", jconf->search_root->pass1.determine_score_thres);
fprintf(fp, " int: frame duration thres (%d)\n", jconf->search_root->pass1.determine_duration_thres);
#endif
fprintf(fp, "\n--- Recognizer / Search Options (-SR) ----------------------------\n");
fprintf(fp, "\n Search Parameters for the First Pass:\n");
fprintf(fp, " [-b beamwidth] beam width (by state num) (guessed)\n");
fprintf(fp, " (0: full search, -1: force guess)\n");
#ifdef WPAIR
# ifdef WPAIR_KEEP_NLIMIT
fprintf(fp, " [-nlimit N] keeps only N tokens on each state (%d)\n", jconf->search_root->pass1.wpair_keep_nlimit);
# endif
#endif
#ifdef SEPARATE_BY_UNIGRAM
fprintf(fp, " [-sepnum wordnum] (n-gram) # of hi-freq word isolated from tree (%d)\n", jconf->lm_root->separate_wnum);
#endif
#ifdef HASH_CACHE_IW
fprintf(fp, " [-iwcache percent] (n-gram) amount of inter-word LM cache (%3d)\n", jconf->search_root->pass1.iw_cache_rate);
#endif
fprintf(fp, " [-1pass] do 1st pass only, omit 2nd pass\n");
fprintf(fp, " [-inactive] recognition process not active on startup\n");
fprintf(fp, "\n Search Parameters for the Second Pass:\n");
fprintf(fp, " [-b2 hyponum] word envelope beam width (by hypo num) (%d)\n",jconf->search_root->pass2.enveloped_bestfirst_width);
fprintf(fp, " [-n N] # of sentence to find (%d)\n", jconf->search_root->pass2.nbest);
fprintf(fp, " [-output N] # of sentence to output (%d)\n",jconf->search_root->output.output_hypo_maxnum);
#ifdef SCAN_BEAM
fprintf(fp, " [-sb score] score beam threshold (by score) (%.1f)\n", jconf->search_root->pass2.scan_beam_thres);
#endif
fprintf(fp, " [-s hyponum] global stack size of hypotheses (%d)\n", jconf->search_root->pass2.stack_size);
fprintf(fp, " [-m hyponum] hypotheses overflow threshold num (%d)\n", jconf->search_root->pass2.hypo_overflow);
fprintf(fp, " [-lookuprange N] frame lookup range in word expansion (%d)\n", jconf->search_root->pass2.lookup_range);
fprintf(fp, " [-looktrellis] (dfa) expand only backtrellis words\n");
fprintf(fp, " [-[no]multigramout] (dfa) output per-grammar results\n");
fprintf(fp, " [-oldtree] (dfa) use old build_wchmm()\n");
#ifdef PASS1_IWCD
fprintf(fp, " [-oldiwcd] (dfa) use full lcdset\n");
#endif
fprintf(fp, " [-iwsp] insert sp for all word end (multipath)(off)\n");
fprintf(fp, " [-iwsppenalty] trans. penalty for iwsp (multipath) (%.1f)\n", jconf->am_root->iwsp_penalty);
fprintf(fp, "\n Short-pause Segmentation:\n");
fprintf(fp, " [-spsegment] enable short-pause segmentation\n");
fprintf(fp, " [-spdur] length threshold of sp frames (%d)\n", jconf->search_root->successive.sp_frame_duration);
#ifdef SPSEGMENT_NAIST
fprintf(fp, " [-spmargin] backstep margin on speech trigger (%d)\n", jconf->search_root->successive.sp_margin);
fprintf(fp, " [-spdelay] delay on speech trigger (%d)\n", jconf->search_root->successive.sp_delay);
#endif
fprintf(fp, " [-pausemodels str] comma-delimited list of pause models for segment\n");
fprintf(fp, "\n Graph Output with graph-oriented search:\n");
fprintf(fp, " [-lattice] enable word graph (lattice) output\n");
fprintf(fp, " [-confnet] enable confusion network output\n");
fprintf(fp, " [-nolattice]][-noconfnet] disable lattice / confnet output\n");
fprintf(fp, " [-graphrange N] merge same words in graph (%d)\n", jconf->search_root->graph.graph_merge_neighbor_range);
fprintf(fp, " -1: not merge, leave same loc. with diff. score\n");
fprintf(fp, " 0: merge same words at same location\n");
fprintf(fp, " >0: merge same words around the margin\n");
#ifdef GRAPHOUT_DEPTHCUT
fprintf(fp, " [-graphcut num] graph cut depth at postprocess (-1: disable)(%d)\n", jconf->search_root->graph.graphout_cut_depth);
#endif
#ifdef GRAPHOUT_LIMIT_BOUNDARY_LOOP
fprintf(fp, " [-graphboundloop num] max. num of boundary adjustment loop (%d)\n", jconf->search_root->graph.graphout_limit_boundary_loop_num);
#endif
#ifdef GRAPHOUT_SEARCH_DELAY_TERMINATION
fprintf(fp, " [-graphsearchdelay] inhibit search termination until 1st sent. found\n");
fprintf(fp, " [-nographsearchdelay] disable it (default)\n");
#endif
fprintf(fp, "\n Forced Alignment:\n");
fprintf(fp, " [-walign] optionally output word alignments\n");
fprintf(fp, " [-palign] optionally output phoneme alignments\n");
fprintf(fp, " [-salign] optionally output state alignments\n");
#ifdef CONFIDENCE_MEASURE
fprintf(fp, "\n Confidence Score:\n");
#ifdef CM_MULTIPLE_ALPHA
fprintf(fp, " [-cmalpha f t s] CM smoothing factor (from, to, step)\n");
#else
fprintf(fp, " [-cmalpha value] CM smoothing factor (%f)\n", jconf->search_root->annotate.cm_alpha);
#endif
#ifdef CM_SEARCH_LIMIT
fprintf(fp, " [-cmthres value] CM threshold to cut hypo on 2nd pass (%f)\n", jconf->search_root->annotate.cm_cut_thres);
#endif
#endif /* CONFIDENCE_MEASURE */
fprintf(fp, "\n Message Output:\n");
fprintf(fp, " [-fallback1pass] use 1st pass result when search failed\n");
fprintf(fp, " [-progout] progressive output in 1st pass\n");
fprintf(fp, " [-proginterval] interval of progout in msec (%d)\n", jconf->search_root->output.progout_interval);
fprintf(fp, "\n-------------------------------------------------\n");
j_jconf_free(jconf);
/* output application-side options */
useropt_show_desc(fp);
}
int
main(int argc, char *argv[])
{
FILE *fp;
Recog *recog;
Jconf *jconf;
/* inihibit system log output (default: stdout) */
//jlog_set_output(NULL);
/* output system log to a file */
// FILE *fp = fopen(logfile, "w"); jlog_set_output(fp);
/* if no option argument, output julius usage and exit */
if (argc == 1) {
fprintf(stderr, "Julius rev.%s - based on ", JULIUS_VERSION);
j_put_version(stderr);
fprintf(stderr, "Try '-setting' for built-in engine configuration.\n");
fprintf(stderr, "Try '-help' for run time options.\n");
return -1;
}
/* add application options */
record_add_option();
module_add_option();
charconv_add_option();
j_add_option("-separatescore", 0, 0, "output AM and LM scores separately", opt_separatescore);
j_add_option("-logfile", 1, 1, "output log to file", opt_logfile);
j_add_option("-nolog", 0, 0, "not output any log", opt_nolog);
j_add_option("-outfile", 0, 0, "save result in separate .out file", opt_outfile);
j_add_option("-help", 0, 0, "display this help", opt_help);
j_add_option("--help", 0, 0, "display this help", opt_help);
/* create a configuration variables container */
jconf = j_jconf_new();
// j_config_load_file(jconf, jconffile);
if (j_config_load_args(jconf, argc, argv) == -1) {
fprintf(stderr, "Try `-help' for more information.\n");
return -1;
}
/* output system log to a file */
if (nolog) {
jlog_set_output(NULL);
} else if (logfile) {
fp = fopen(logfile, "w");
jlog_set_output(fp);
}
/* here you can set/modify any parameter in the jconf before setup */
// jconf->input.input_speech = SP_MIC;
/* Fixate jconf parameters: it checks whether the jconf parameters
are suitable for recognition or not, and set some internal
parameters according to the values for recognition. Modifying
a value in jconf after this function may be errorous.
*/
if (j_jconf_finalize(jconf) == FALSE) {
if (logfile) fclose(fp);
return -1;
}
/* create a recognition instance */
recog = j_recog_new();
/* assign configuration to the instance */
recog->jconf = jconf;
/* load all files according to the configurations */
if (j_load_all(recog, jconf) == FALSE) {
fprintf(stderr, "ERROR: Error in loading model\n");
if (logfile) fclose(fp);
return -1;
}
#ifdef USER_LM_TEST
{
PROCESS_LM *lm;
for(lm=recog->lmlist;lm;lm=lm->next) {
if (lm->lmtype == LM_PROB) {
j_regist_user_lm_func(lm, my_uni, my_bi, my_lm);
}
}
#endif
/* checkout for recognition: build lexicon tree, allocate cache */
if (j_final_fusion(recog) == FALSE) {
fprintf(stderr, "ERROR: Error while setup work area for recognition\n");
j_recog_free(recog);
if (logfile) fclose(fp);
return -1;
}
/* Set up some application functions */
/* set character conversion mode */
if (charconv_setup() == FALSE) {
if (logfile) fclose(fp);
return -1;
}
if (is_module_mode()) {
/* set up for module mode */
/* register result output callback functions to network module */
module_setup(recog, NULL);
} else {
/* register result output callback functions to stdout */
setup_output_tty(recog, NULL);
}
/* if -outfile option specified, callbacks for file output will be
regitered */
if (outfile_enabled) {
if (jconf->input.speech_input == SP_MFCFILE || jconf->input.speech_input == SP_RAWFILE) {
setup_output_file(recog, NULL);
} else {
fprintf(stderr, "Warning: -outfile works only for file input, disabled now\n");
outfile_enabled = FALSE;
}
}
/* setup recording if option was specified */
record_setup(recog, NULL);
/* on module connect with client */
if (is_module_mode()) module_server();
/* initialize and standby the specified audio input source */
/* for microphone or other threaded input, ad-in thread starts here */
if (j_adin_init(recog) == FALSE) return;
/* output system information to log */
j_recog_info(recog);
#ifdef VISUALIZE
/* Visualize: initialize GTK */
visual_init(recog);
callback_add(recog, CALLBACK_EVENT_RECOGNITION_END, visual_show, NULL);
callback_add(recog, CALLBACK_EVENT_PASS2_BEGIN, visual2_init, NULL);
callback_add(recog, CALLBACK_DEBUG_PASS2_POP, visual2_popped, NULL);
callback_add(recog, CALLBACK_DEBUG_PASS2_PUSH, visual2_next_word, NULL);
/* below should be called at result */
visual2_best(now, winfo);
/* 音声取り込みはコールバックで新規作成 */
/* 第2パスで認識結果出力時に以下を実行 */
visual2_best(now, recog->model->winfo);
#endif
/* if no grammar specified on startup, start with pause status */
{
RecogProcess *r;
boolean ok_p;
ok_p = TRUE;
for(r=recog->process_list;r;r=r->next) {
if (r->lmtype == LM_DFA) {
if (r->lm->winfo == NULL) { /* stop when no grammar found */
j_request_pause(recog);
}
}
}
}
/* enter recongnition loop */
main_recognition_stream_loop(recog);
/* end proc */
if (is_module_mode()) module_disconnect();
/* release all */
j_recog_free(recog);
if (logfile) fclose(fp);
return(0);
}
/**
* <JA>
* メイン関数
*
* @param argc [in] 引数列の長さ
* @param argv [in] 引数列
*
* @return
* </JA>エラー時 1，通常終了時 0 を返す．
* <EN>
* Main function.
*
* @param argc [in] number of argument.
* @param argv [in] array of arguments.
*
* @return 1 on error, 0 on success.
* </EN>
*/
int
main(int argc, char *argv[])
{
Recog *recog;
Jconf *jconf;
/* create instance */
jconf = j_jconf_new();
/* register application options */
j_add_option("-freq", 1, 1, "sampling frequency in Hz", opt_freq);
j_add_option("-raw", 0, 0, "save in raw (BE) format", opt_raw);
j_add_option("-h", 0, 0, "display this help", opt_help);
j_add_option("-help", 0, 0, "display this help", opt_help);
j_add_option("--help", 0, 0, "display this help", opt_help);
/* when no argument, output help and exit */
if (argc <= 1) {
opt_help(jconf, NULL, 0);
return 0;
}
/* regard last arg as filename */
if (strmatch(argv[argc-1], "-")) {
stout = TRUE;
use_raw = TRUE;
} else {
filename = argv[argc-1];
}
/* set default as same as "-input mic" */
jconf->input.type = INPUT_WAVEFORM;
jconf->input.speech_input = SP_MIC;
jconf->input.device = SP_INPUT_DEFAULT;
/* read arguments and set parameters */
if (j_config_load_args(jconf, argc-1, argv) == -1) {
fprintf(stderr, "Error reading arguments\n");
return -1;
}
/* exit if no file name specified */
if (filename == NULL && stout == FALSE) {
opt_help(jconf, NULL, 0);
return -1;
}
/* finalize config */
//if (j_jconf_finalize(jconf) == FALSE) return -1;
/* set Julius default parameters for unspecified acoustic parameters */
apply_para(&(jconf->am_root->analysis.para), &(jconf->am_root->analysis.para_default));
/* set some values */
jconf->input.sfreq = jconf->am_root->analysis.para.smp_freq;
jconf->input.period = jconf->am_root->analysis.para.smp_period;
jconf->input.frameshift = jconf->am_root->analysis.para.frameshift;
jconf->input.framesize = jconf->am_root->analysis.para.framesize;
/* preliminary check of output file */
/* (output file will be opened later when input is triggered) */
if (!stout) {
if (access(filename, F_OK) == 0) {
if (access(filename, W_OK) == 0) {
fprintf(stderr, "Warning: overwriting file \"%s\"\n", filename);
} else {
perror("adinrec");
return(1);
}
}
}
/* set signal handlers to properly close output file */
if (signal(SIGINT, interrupt_record) == SIG_ERR) {
fprintf(stderr, "Warning: signal intterupt may collapse output\n");
}
if (signal(SIGTERM, interrupt_record) == SIG_ERR) {
fprintf(stderr, "Warning: signal intterupt may collapse output\n");
}
#ifdef SIGPIPE
if (signal(SIGPIPE, interrupt_record) == SIG_ERR) {
fprintf(stderr, "Warning: signal intterupt may collapse output\n");
}
#endif
#ifdef SIGQUIT
if (signal(SIGQUIT, interrupt_record) == SIG_ERR) {
fprintf(stderr, "Warning: signal intterupt may collapse output\n");
}
#endif
recog = j_recog_new();
recog->jconf = jconf;
/* initialize input device */
if (j_adin_init(recog) == FALSE) {
fprintf(stderr, "Error in initializing adin device\n");
return -1;
}
/* open device */
if (j_open_stream(recog, NULL) < 0) {
fprintf(stderr, "Error in opening adin device\n");
}
/* do recoding */
speechlen = 0;
sfreq = recog->jconf->input.sfreq;
fprintf(stderr, "<<< please speak >>>"); /* moved from adin-cut.c */
adin_go(adin_callback_file, NULL, recog);
/* close device */
adin_end(recog->adin);
/* close output file */
close_file();
return 0;
}
