bool SpRecog::start(Application *app)
{
if (j_adin_init(this->recog) == FALSE) {
app->tell("error in j_adin_init");
return false;
}
j_recog_info(this->recog); /* output system information to log */
switch(j_open_stream(this->recog, NULL)) {
case 0:
app->tell("ok j_open_stream");
break;
case -1:
app->tell("error in input stream");
return false;
case -2:
app->tell("error in beginning input");
return false;
}
fflush(this->srm_log_fp);
int ret = j_recognize_stream(this->recog);
if (ret == -1) {
app->tell("error j_recognize_stream");
return false;
}
return true;
}
void SpeechSystem::setup()
{
// ewwww
jconf = j_config_load_file_new(const_cast<char*>(jconf_filename.c_str()));
/* 2. create recognition instance according to the jconf */
/* it loads models, setup final parameters, build lexicon
and set up work area for recognition */
recog = j_create_instance_from_jconf(jconf);
if (recog == NULL)
{
fprintf(stderr, "Error in startup\n");
return;
}
/*********************/
/* Register callback */
/*********************/
/* register result callback functions */
callback_add(recog, CALLBACK_EVENT_SPEECH_READY, recready, this);
callback_add(recog, CALLBACK_EVENT_SPEECH_START, recstart, this);
callback_add(recog, CALLBACK_RESULT, recdone, this);
/**************************/
/* Initialize audio input */
/**************************/
/* initialize audio input device */
/* ad-in thread starts at this time for microphone */
if (j_adin_init(recog) == FALSE) { /* error */
return;
}
//#ifdef JULIUS_DEBUG
/* output system information to log */
j_recog_info(recog);
//#endif
/***********************************/
/* Open input stream and recognize */
/***********************************/
/* raw speech input (microphone etc.) */
switch(j_open_stream(recog, NULL)) {
case 0: /* succeeded */
break;
case -1: /* error */
fprintf(stderr, "error in input stream\n");
return;
case -2: /* end of recognition process */
fprintf(stderr, "failed to begin input stream\n");
return;
}
startThread(true, false); // blocking, verbose
}
/**
* <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;
}
/**
* <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;
}
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);
}
int main(int argc, char* argv[])
{
// Jconf: configuration parameters
// load configurations from command arguments
Jconf *jconf = j_config_load_args_new(argc, argv);
if (jconf == NULL) {
std::cout << "Error @ j_config_load_args_new" << std::endl;
return -1;
}
// Recog: Top level instance for the whole recognition process
// create recognition instance according to the jconf
Recog *recog = j_create_instance_from_jconf(jconf);
if (recog == NULL) {
std::cout << "Error @ j_create_instance_from_jconf" << std::endl;
return -1;
}
// Regster callback
callback_add(recog, CALLBACK_EVENT_SPEECH_READY, [](Recog *recog, void*) {
std::cout << "<<< PLEASE SPEAK! >>>" << std::endl;
}, NULL);
callback_add(recog, CALLBACK_EVENT_SPEECH_START, [](Recog *recog, void*) {
std::cout << "...SPEECH START..." << std::endl;
}, NULL);
callback_add(recog, CALLBACK_RESULT, [](Recog *recog, void*) {
for (const RecogProcess *r = recog->process_list; r; r = r->next) {
WORD_INFO *winfo = r->lm->winfo;
for (int n = 0; n < r->result.sentnum; ++n) {
Sentence *s = &(r->result.sent[n]);
WORD_ID *seq = s->word;
int seqnum = s->word_num;
for (int i = 0; i < seqnum; ++i) {
std::cout << winfo->woutput[seq[i]];
}
}
}
}, NULL);
// Initialize audio input
if (j_adin_init(recog) == FALSE) {
return -1;
}
// output system information to log
j_recog_info(recog);
// Open input stream and recognize
switch (j_open_stream(recog, NULL)) {
case 0: break; // success
case -1: std::cout << "Error in input stream" << std::endl; return -1;
case -2: std::cout << "Failed to begin input stream" << std::endl; return -1;
}
// Recognition loop
int ret = j_recognize_stream(recog);
if (ret == -1) return -1;
// exit
j_close_stream(recog);
j_recog_free(recog);
return 0;
}
//========================
// Create engine instance
//========================
bool cJulius::createEngine( void )
{
#ifdef APP_ADIN
ADIn *a;
#endif
if (!m_jconf) return false;
if (m_recog) return false;
#ifdef APP_ADIN
if (m_appsource != 0) {
switch(m_appsource) {
case 1: // buffer input, batch
m_recog->jconf->input.type = INPUT_WAVEFORM;
m_recog->jconf->input.speech_input = SP_RAWFILE;
m_recog->jconf->decodeopt.realtime_flag = FALSE;
break;
case 2: // buffer input, incremental
m_recog->jconf->input.type = INPUT_WAVEFORM;
m_recog->jconf->input.speech_input = SP_RAWFILE;
m_recog->jconf->decodeopt.realtime_flag = TRUE;
break;
}
}
#endif
// Create engine instance
m_recog = j_create_instance_from_jconf(m_jconf);
if (m_recog == NULL) {
return false;
}
// Register callbacks
callback_add(m_recog, CALLBACK_EVENT_PROCESS_ONLINE, ::callback_engine_active, this);
callback_add(m_recog, CALLBACK_EVENT_PROCESS_OFFLINE, ::callback_engine_inactive, this);
callback_add(m_recog, CALLBACK_EVENT_SPEECH_READY, ::callback_audio_ready, this);
callback_add(m_recog, CALLBACK_EVENT_SPEECH_START, ::callback_audio_begin, this);
callback_add(m_recog, CALLBACK_EVENT_SPEECH_STOP, ::callback_audio_end, this);
callback_add(m_recog, CALLBACK_EVENT_RECOGNITION_BEGIN, ::callback_recog_begin, this);
callback_add(m_recog, CALLBACK_EVENT_RECOGNITION_END, ::callback_recog_end, this);
callback_add(m_recog, CALLBACK_EVENT_PASS1_FRAME, ::callback_recog_frame, this);
callback_add(m_recog, CALLBACK_EVENT_PAUSE, ::callback_engine_pause, this);
callback_add(m_recog, CALLBACK_EVENT_RESUME, ::callback_engine_resume, this);
callback_add(m_recog, CALLBACK_RESULT, ::callback_result_final, this);
callback_add(m_recog, CALLBACK_PAUSE_FUNCTION, ::callback_wait_for_resume, this);
#ifdef APP_ADIN
// Initialize application side audio input
if (m_appsource != 0) {
a = m_recog->adin;
switch(m_appsource) {
case 1: // buffer input, batch
a->ad_standby = NULL;
a->ad_begin = NULL;
a->ad_end = NULL;
a->ad_resume = NULL;
a->ad_pause = NULL;
a->ad_terminate = NULL;
a->ad_read = callback_adin_fetch_input;
a->ad_input_name = NULL;
a->silence_cut_default = FALSE;
a->enable_thread = FALSE;
break;
case 2: // buffer input, incremental
a->ad_standby = NULL;
a->ad_begin = NULL;
a->ad_end = NULL;
a->ad_resume = NULL;
a->ad_pause = NULL;
a->ad_terminate = NULL;
a->ad_read = callback_adin_fetch_input;
a->ad_input_name = NULL;
a->silence_cut_default = FALSE;
a->enable_thread = FALSE;
break;
}
a->ds = NULL;
a->down_sample = FALSE;
if (adin_standby(a, m_recog->jconf->input.sfreq, NULL) == FALSE) return false;
if (adin_setup_param(a, m_recog->jconf) == FALSE) return false;
a->input_side_segment = FALSE;
} else {
// Let JuliusLib get audio input
if (! j_adin_init( m_recog ) ) return false;
}
#else
if (! j_adin_init( m_recog ) ) return false;
#endif
return true;
}
xreturn::r<bool> JuliusPlus::JuliusFileStart()
{
assert(this->recogFile == NULL);
assert(this->jconfFile == NULL);
const char* argv[]={
"juliusplus"
,"-C"
,"testfile.jconf"
};
int argc = sizeof(argv)/sizeof(argv[0]);
//julius は C関数だから、const外して char** にするしかない。
this->jconfFile = j_config_load_args_new(argc, (char**)argv);
/* else, you can load configurations from a jconf file */
//jconf = j_config_load_file_new(jconf_filename);
if (this->jconfFile == NULL)
{
return xreturn::error("Try `-help' for more information.\n");
}
/* 2. create recognition instance according to the jconf */
/* it loads models, setup final parameters, build lexicon
and set up work area for recognition */
this->recogFile = j_create_instance_from_jconf(this->jconfFile);
if (this->recogFile == NULL)
{
return xreturn::error("Error in startup(j_create_instance_from_jconf)\n");
}
struct _ref{
static void output_result(Recog *recog, void *_this)
{
((JuliusPlus*)_this)->OnOutputResultFile(recog);
}
};
callback_add(this->recogFile, CALLBACK_RESULT, _ref::output_result, this);
// Initialize audio input
if (j_adin_init(this->recogFile) == FALSE)
{
return xreturn::error("Error in startup(j_adin_init)\n");
}
//以上、準備だけしておいて、
//認識ルーチンは、後から呼びます。
/*
int ret = j_open_stream(recogFile, "nano.wav");
if(ret < 0)
{
return xreturn::error("Error in startup(j_open_stream)\n");
}
j_recognize_stream(recogFile);
ret = j_open_stream(recogFile, "nano.wav");
if(ret < 0)
{
return xreturn::error("Error in startup(j_open_stream)\n");
}
j_recognize_stream(recogFile);
*/
return true;
}
xreturn::r<bool> JuliusPlus::JuliusStart()
{
assert(this->recog == NULL);
assert(this->jconf == NULL);
assert(this->Thread == NULL);
const char* argv[]={
"juliusplus"
,"-C"
,"testmic.jconf"
};
int argc = sizeof(argv)/sizeof(argv[0]);
int ret;
//julusはC関数なので、const外して char** にするしかない・・・
this->jconf = j_config_load_args_new(argc, (char**)argv);
/* else, you can load configurations from a jconf file */
//jconf = j_config_load_file_new(jconf_filename);
if (this->jconf == NULL)
{
return xreturn::error("Try `-help' for more information.\n");
}
/* 2. create recognition instance according to the jconf */
/* it loads models, setup final parameters, build lexicon
and set up work area for recognition */
this->recog = j_create_instance_from_jconf(this->jconf);
if (this->recog == NULL)
{
return xreturn::error("Error in startup(j_create_instance_from_jconf)\n");
}
struct _ref{
static void status_recready(Recog *recog, void *_this)
{
((JuliusPlus*)_this)->OnStatusRecready(recog);
}
static void status_recstart(Recog *recog, void *_this)
{
((JuliusPlus*)_this)->OnStatusRecstart(recog);
}
static void output_result(Recog *recog, void *_this)
{
((JuliusPlus*)_this)->OnOutputResult(recog);
}
static void record_adin_trigger(Recog *recog, SP16 *speech, int samplenum, void *_this)
{
((JuliusPlus*)_this)->OnRecordAdinTrigger(recog,speech,samplenum);
}
};
callback_add(this->recog, CALLBACK_EVENT_SPEECH_READY, _ref::status_recready, this);
callback_add(this->recog, CALLBACK_EVENT_SPEECH_START, _ref::status_recstart, this);
callback_add(this->recog, CALLBACK_RESULT, _ref::output_result, this);
callback_add_adin(this->recog, CALLBACK_ADIN_TRIGGERED, _ref::record_adin_trigger, this);
// Initialize audio input
if (j_adin_init(this->recog) == FALSE)
{
return xreturn::error("Error in startup(j_adin_init)\n");
}
//output system information to log
//j_recog_info(this->recog);
ret = j_open_stream(recog, NULL);
if(ret < 0)
{
return xreturn::error("Error in startup(j_open_stream)\n");
}
this->Thread = new boost::thread( [&]()
{
j_recognize_stream(recog);
} );
return true;
}
/* Julius_Thread::run: main loop */
void Julius_Thread::run()
{
char *tmp;
char buff[MMDAGENT_MAXBUFLEN];
FILE *fp;
if(m_jconf != NULL || m_recog != NULL || m_mmdagent == NULL || m_thread < 0 || m_languageModel == 0 || m_dictionary == 0 || m_acousticModel == 0 || m_triphoneList == 0 || m_configFile == 0)
return;
/* set latency */
sprintf(buff, "PA_MIN_LATENCY_MSEC=%d", JULIUSTHREAD_LATENCY);
putenv(buff);
sprintf(buff, "LATENCY_MSEC=%d", JULIUSTHREAD_LATENCY);
putenv(buff);
/* turn off log */
jlog_set_output(NULL);
/* load models */
tmp = MMDAgent_pathdup(m_languageModel);
sprintf(buff, "-d \"%s\"", tmp);
free(tmp);
m_jconf = j_config_load_string_new(buff);
if (m_jconf == NULL) {
return;
}
tmp = MMDAgent_pathdup(m_dictionary);
sprintf(buff, "-v \"%s\"", tmp);
free(tmp);
if(j_config_load_string(m_jconf, buff) < 0) {
return;
}
tmp = MMDAgent_pathdup(m_acousticModel);
sprintf(buff, "-h \"%s\"", tmp);
free(tmp);
if(j_config_load_string(m_jconf, buff) < 0) {
return;
}
tmp = MMDAgent_pathdup(m_triphoneList);
sprintf(buff, "-hlist \"%s\"", tmp);
free(tmp);
if(j_config_load_string(m_jconf, buff) < 0) {
return;
}
/* load config file */
tmp = MMDAgent_pathdup(m_configFile);
if(j_config_load_file(m_jconf, tmp) < 0) {
free(tmp);
return;
}
free(tmp);
/* load user dictionary */
fp = MMDAgent_fopen(m_userDictionary, "r");
if(fp != NULL) {
fclose(fp);
tmp = MMDAgent_pathdup(m_userDictionary);
j_add_dict(m_jconf->lm_root, tmp);
free(tmp);
}
/* create instance */
m_recog = j_create_instance_from_jconf(m_jconf);
if (m_recog == NULL) {
return;
}
/* register callback functions */
callback_add(m_recog, CALLBACK_EVENT_RECOGNITION_BEGIN, callbackRecogBegin, this);
callback_add(m_recog, CALLBACK_RESULT, callbackRecogResult, this);
if (!j_adin_init(m_recog)) {
return;
}
if (j_open_stream(m_recog, NULL) != 0) {
return;
}
/* setup logger */
m_logger.setup(m_recog);
/* start logger */
m_logger.setActiveFlag(true);
/* start recognize */
j_recognize_stream(m_recog);
}
