galat****@lists*****
galat****@lists*****
2006年 10月 19日 (木) 12:26:33 JST
Index: gtalk/do_output.c
diff -u gtalk/do_output.c:1.19 gtalk/do_output.c:1.20
--- gtalk/do_output.c:1.19 Thu Sep 14 22:01:25 2006
+++ gtalk/do_output.c Thu Oct 19 12:26:33 2006
@@ -8,7 +8,8 @@
/* Keiichi Tokuda, Takayoshi Yoshimura, Heiga Zen */
/* (Nagoya Institute of Technology) */
/* All rights reserved */
-
+/* */
+/* $Id: do_output.c,v 1.20 2006/10/19 03:26:33 sako Exp $ */
/*
波形を生成しながらの音声出力処理は pthread を用いて実装されている。
これは AUTO_DA が定義されていれば，組み込まれる。
@@ -39,6 +40,7 @@
void inqSpeakUtt();
void inqSpeakStat();
void do_output_file(char *);
+void abort_auto_output();
#ifdef USE_SPLIB
#include "do_output_sp.c"
@@ -58,6 +60,8 @@
int org_vol, org_channels, org_precision, org_freq;
int forced_stereo;
#endif /* LINUX */
+int current_pos;
+int prev_tell_pos_ms;
#ifdef SOLARIS
audio_info_t org_data;
@@ -113,7 +117,11 @@
short *out;
{
int i;
+ int pos, pos_ms, interval_ms;
+ int samp_rate;
+ samp_rate = data_type[DTYPE].sample;
+
// fwrite( out, sizeof(short), leng, adfp);
for( i=0; i<leng; ++i ) {
fwrite( &(out[i]), sizeof(short), 1, adfp);
@@ -121,9 +129,38 @@
if( forced_stereo ) {
fwrite( &(out[i]), sizeof(short), 1, adfp);
}
+
+ if(current_pos % 128 == 0) {
+#ifdef LINUX
+ count_info info;
+
+ if( ioctl( ADFD, SNDCTL_DSP_GETOPTR, &info ) != -1 ) {
+ pos = info.bytes / 2;
+ if( forced_stereo ) {
+ pos /= 2;
+ }
+ } else {
+ pos = current_pos;
+ }
+#else
+ pos = current_pos;
+#endif
+ pos_ms = (1000 * pos) / samp_rate;
+ interval_ms = pos_ms - prev_tell_pos_ms;
+
+ if( slot_Speak_syncinterval > 0 && interval_ms >= slot_Speak_syncinterval ) {
+ RepMsg("tell Speak.sync = %d\n", pos_ms);
+ /*prev_tell_pos_ms = pos_ms;*/
+ while( prev_tell_pos_ms + slot_Speak_syncinterval <= pos_ms ) {
+ prev_tell_pos_ms += slot_Speak_syncinterval;
+ }
+ }
+ }
+ ++current_pos;
}
write( ADFD, out, 0);
fflush( adfp);
+
}
void init_audiodev(dtype)
@@ -184,6 +221,8 @@
ioctl(ADFD,AUDIO_SETINFO,&data);
#endif /* SOLARIS */
+ current_pos = 0;
+ prev_tell_pos_ms = 0;
}
void reset_audiodev()
@@ -206,6 +245,8 @@
ioctl( ACFD, AUDIO_SETINFO, &org_data);
close( ACFD);
#endif /* SOLARIS */
+ current_pos = 0;
+ prev_tell_pos_ms = 0;
}
/*---------------------------------------------------------------------*/
Index: gtalk/do_output_sp.c
diff -u gtalk/do_output_sp.c:1.3 gtalk/do_output_sp.c:1.4
--- gtalk/do_output_sp.c:1.3 Sat Sep 2 09:58:09 2006
+++ gtalk/do_output_sp.c Thu Oct 19 12:26:33 2006
@@ -1,18 +1,22 @@
+/* $Id: do_output_sp.c,v 1.4 2006/10/19 03:26:33 sako Exp $ */
#include <sp/spBaseLib.h>
#include <sp/spAudioLib.h>
#define GTALK_AUDIO_BLOCKSIZE 1024
+#define GTALK_USE_SP_PLUGIN_OUTPUT
+int RepMsg(char *, ...);
void abort_auto_output();
static volatile spAudio gtalk_sp_audio = NULL;
static volatile int gtalk_aborted_flag = 0;
-
-int in_auto_play;
+static volatile long gtalk_prev_tell_pos_ms = 0;
/* nonzero value restarts play from beginning on replay */
static int gtalk_replay_init_flag = 0;
+int in_auto_play;
+
void output_speaker(int total) {}
void init_output() {}
@@ -21,16 +25,28 @@
void set_da_signal() {}
-void abort_demanded_output() {
- if (gtalk_sp_audio != NULL) {
- long position;
+static int get_current_pos(spAudio audio)
+{
+ int pos_ms;
+ long position;
- if (spGetAudioOutputPosition(gtalk_sp_audio, &position) == SP_TRUE) {
- talked_DA_msec = (1000 * position) / SAMPLE_RATE;
- spDebug(1, "abort_output",
+ pos_ms = -1;
+
+ if (audio != NULL) {
+ if (spGetAudioOutputPosition(audio, &position) == SP_TRUE) {
+ pos_ms = (1000 * position) / SAMPLE_RATE;
+ spDebug(80, "get_current_pos",
"spGetAudioOutputPosition: position = %ld, da_msec = %d, SAMPLE_RATE = %d\n",
- position, talked_DA_msec, SAMPLE_RATE);
+ position, pos_ms, SAMPLE_RATE);
}
+ }
+
+ return pos_ms;
+}
+
+void abort_demanded_output() {
+ if (gtalk_sp_audio != NULL) {
+ talked_DA_msec = get_current_pos(gtalk_sp_audio);
if (!gtalk_aborted_flag) {
spDebug(1, "abort_output", "output aborted\n");
@@ -44,6 +60,32 @@
}
+spBool output_pos_func(spAudio audio, spAudioCallbackType call_type,
+ void *data1, void *data2, void *user_data)
+{
+ long pos_ms;
+ long *pos;
+ long current_interval;
+
+ if (call_type == SP_AUDIO_OUTPUT_POSITION_CALLBACK) {
+ pos = (long *)data1;
+ pos_ms = 1000 * (*pos) / SAMPLE_RATE;
+ current_interval = pos_ms - gtalk_prev_tell_pos_ms;
+ spDebug(100, "output_pos_func",
+ "pos = %ld, pos_ms = %ld, slot_Speak_syncinterval = %d, current_interval = %ld\n",
+ *pos, pos_ms, slot_Speak_syncinterval, current_interval);
+
+ if (slot_Speak_syncinterval > 0 && current_interval >= (long)slot_Speak_syncinterval) {
+ if (!gtalk_aborted_flag) {
+ RepMsg("tell Speak.sync = %ld\n", pos_ms);
+ }
+ gtalk_prev_tell_pos_ms = pos_ms;
+ }
+ }
+
+ return SP_TRUE;
+}
+
static spThreadReturn do_output_thread(void *data)
{
int total;
@@ -55,10 +97,16 @@
total = *(int *)data;
+ if (in_auto_play && slot_Auto_play_delay > 0) {
+ spMSleep(slot_Auto_play_delay);
+ }
+
if (gtalk_sp_audio == NULL) {
gtalk_sp_audio = spInitAudio();
spSetAudioSampleRate(gtalk_sp_audio, (double)SAMPLE_RATE);
spSetAudioBufferSize(gtalk_sp_audio, GTALK_AUDIO_BLOCKSIZE);
+ spSetAudioCallbackFunc(gtalk_sp_audio, SP_AUDIO_OUTPUT_POSITION_CALLBACK,
+ output_pos_func, NULL);
}
/* open "write only" mode */
@@ -70,6 +118,8 @@
offset = 0;
block_length = GTALK_AUDIO_BLOCKSIZE * sizeof(short);
+ gtalk_prev_tell_pos_ms = 0;
+
while (!gtalk_aborted_flag && offset < total) {
spDebug(80, "do_output_thread", "offset = %ld, total = %ld\n", offset, total);
@@ -79,18 +129,13 @@
offset += block_length;
}
- if (!gtalk_aborted_flag) {
- long position;
+ spCloseAudioDevice(gtalk_sp_audio);
- if (spGetAudioOutputPosition(gtalk_sp_audio, &position) == SP_TRUE) {
- talked_DA_msec = (1000 * position) / SAMPLE_RATE;
- }
+ if (!gtalk_aborted_flag) {
if ( prop_Speak_len == AutoOutput ) inqSpeakLen();
if ( prop_Speak_utt == AutoOutput ) inqSpeakUtt();
}
- spCloseAudioDevice(gtalk_sp_audio);
-
strcpy( slot_Speak_stat, "IDLE" );
if( prop_Speak_stat == AutoOutput ) inqSpeakStat();
@@ -103,7 +148,7 @@
void do_output_file_sp(char *sfile)
{
-#if 1
+#ifdef GTALK_USE_SP_PLUGIN_OUTPUT
static char o_plugin_name[SP_MAX_LINE];
spPlugin *o_plugin;
spWaveInfo wave_info;
@@ -132,50 +177,46 @@
return;
}
-void do_output(char *fn)
+void do_output_da()
{
static int total;
-
- in_auto_play = 0;
+ static void *thread = NULL;
/*total = SAMPLE_RATE * FRAME_RATE * (totalframe - 1) / 1000;*/
total = wave.nsample;
- if(fn == NULL){
- static void *thread = NULL;
- talked_DA_msec = -1;
- already_talked = 1;
+ talked_DA_msec = -1;
+ already_talked = 1;
- if (thread != NULL) {
- if (gtalk_sp_audio != NULL && gtalk_replay_init_flag) {
- spStopAudio(gtalk_sp_audio);
- }
-
- spWaitThread(thread);
- spDestroyThread(thread);
+ if (thread != NULL) {
+ if (gtalk_sp_audio != NULL && gtalk_replay_init_flag) {
+ spStopAudio(gtalk_sp_audio);
}
- spDebug(1, "do_output", "creating thread...\n");
- gtalk_aborted_flag = 0;
+ spWaitThread(thread);
+ spDestroyThread(thread);
+ }
+ spDebug(1, "do_output_da", "creating thread...\n");
- if ((thread = spCreateThread(0, SP_THREAD_PRIORITY_NORMAL,
- do_output_thread, (void *)&total)) == NULL) {
- spDebug(1, "do_output", "Can't create audio thread\n");
- return;
- }
+ gtalk_aborted_flag = 0;
- spDebug(1, "do_output", "creating thread done\n");
+ if ((thread = spCreateThread(0, SP_THREAD_PRIORITY_NORMAL,
+ do_output_thread, (void *)&total)) == NULL) {
+ spDebug(1, "do_output", "Can't create audio thread\n");
+ return;
+ }
+
+ spDebug(1, "do_output_da", "creating thread done\n");
+}
+
+void do_output(char *fn)
+{
+ in_auto_play = 0;
+
+ if(fn == NULL){
+ do_output_da();
} else {
-#if 0
- FILE *fp;
-
- if ((fp = fopen(fn, "wb")) != NULL) {
- fwrite(wave.data, sizeof (short), total, fp);
- fclose(fp);
- }
-#else
do_output_file_sp( fn );
-#endif
}
}
@@ -197,62 +238,15 @@
--------------------------------------------------------------------*/
#ifdef AUTO_DA
-
void do_auto_output()
{
- static int total;
- static void *thread = NULL;
-
- in_auto_play = 1;
-
- /*total = SAMPLE_RATE * FRAME_RATE * (totalframe - 1) / 1000;*/
- total = wave.nsample;
-
- talked_DA_msec = -1;
- already_talked = 1;
-
- if (thread != NULL) {
- if (gtalk_sp_audio != NULL && gtalk_replay_init_flag) {
- spStopAudio(gtalk_sp_audio);
- }
+ in_auto_play = 1;
- spWaitThread(thread);
- spDestroyThread(thread);
- }
- spDebug(1, "do_output", "creating thread...\n");
-
- gtalk_aborted_flag = 0;
-
- if ((thread = spCreateThread(0, SP_THREAD_PRIORITY_NORMAL,
- do_output_thread, (void *)&total)) == NULL) {
- spDebug(1, "do_output", "Can't create audio thread\n");
- return;
- }
-
- spDebug(1, "do_output", "creating thread done\n");
+ do_output_da();
}
void abort_auto_output() {
- if (gtalk_sp_audio != NULL) {
- long position;
-
- if (spGetAudioOutputPosition(gtalk_sp_audio, &position) == SP_TRUE) {
- talked_DA_msec = (1000 * position) / SAMPLE_RATE;
- spDebug(1, "abort_output",
- "spGetAudioOutputPosition: position = %ld, da_msec = %d, SAMPLE_RATE = %d\n",
- position, talked_DA_msec, SAMPLE_RATE);
- }
-
- if (!gtalk_aborted_flag) {
- spDebug(1, "abort_output", "output aborted\n");
- spStopAudio(gtalk_sp_audio);
- gtalk_aborted_flag = 1;
- }
-
- if ( prop_Speak_len == AutoOutput ) inqSpeakLen();
- if ( prop_Speak_utt == AutoOutput ) inqSpeakUtt();
- }
-
+ abort_demanded_output();
}
#endif /* AUTO_DA */
Index: gtalk/main.c
diff -u gtalk/main.c:1.28 gtalk/main.c:1.29
--- gtalk/main.c:1.28 Fri Oct 13 18:44:44 2006
+++ gtalk/main.c Thu Oct 19 12:26:33 2006
@@ -1,8 +1,10 @@
/* Copyright (c) 2003-2006 */
/* Interactive Speech Technology Consortium (ISTC) */
/* All rights reserved */
-
+/* */
/* The code is developed by Yamashita-lab, Ritsumeikan University */
+/* */
+/* $Id: main.c,v 1.29 2006/10/19 03:26:33 sako Exp $ */
#include <stdio.h>
#include <stdlib.h>
@@ -121,6 +123,7 @@
SlotProp prop_Speak_utt;
SlotProp prop_Speak_len;
SlotProp prop_Speak_stat;
+SlotProp prop_Speak_syncinterval;
/* slots */
char slot_Run[20];
@@ -146,6 +149,7 @@
int slot_Log_text;
int slot_Log_arranged_text;
int slot_Log_sentence;
+int slot_Speak_syncinterval;
/* chaone */
#ifdef WIN32
@@ -219,6 +223,7 @@
int server_getline ( char *buf, int buf_size );
void server_destroy ( void );
/********↑***********************/
+void SetRun( char *rel, char *val);
extern FILE *fp_err;
@@ -242,6 +247,7 @@
prop_Speak_utt = AutoOutput;
prop_Speak_len = AutoOutput;
prop_Speak_stat = AutoOutput;
+ prop_Speak_syncinterval = AutoOutput;
}
/* 初期化: プログラム起動時に一度だけ実行 */
@@ -282,6 +288,7 @@
slot_Speech_file[0] = '\0';
slot_Pros_file[0] = '\0';
prosBuf.nPhoneme = 0;
+ slot_Speak_syncinterval = 1000;
}
void refresh_prosBuf()
@@ -496,6 +503,11 @@
RepMsg( "rep Speak.stat = %s\n", slot_Speak_stat );
}
+void inqSpeakSyncinterval()
+{
+ RepMsg( "rep Speak.syncinterval = %d\n", slot_Speak_syncinterval );
+}
+
/*---------------------------------------------------------*/
/* set command */
/*---------------------------------------------------------*/
@@ -632,6 +644,20 @@
*/
}
+void setSpeakSyncinterval( char *rel, char *val )
+{
+ int interval;
+
+ if( strcmp(rel,"=")!=0 ) { unknown_com(); return; }
+
+ interval = atoi( val );
+
+ if( interval >= 0) {
+ slot_Speak_syncinterval = interval;
+ if( prop_Speak_syncinterval == AutoOutput ) inqSpeakSyncinterval();
+ }
+}
+
/*-------------------*/
void setSave( char *rel, char *filename )
@@ -945,6 +971,7 @@
case S_SpeechFile: setSpeechFile( v_arg[2], v_arg[3] ); break;
case S_ProsFile: setProsFile( v_arg[2], v_arg[3] ); break;
case S_ParsedText: setParsedText( v_arg[2], v_arg[3] ); break;
+ case S_Speak_syncinterval: setSpeakSyncinterval( v_arg[2], v_arg[3] ); break;
case S_AutoPlay:
slot_Auto_play = setLogYesNo( v_arg[2], v_arg[3] ); break;
case S_AutoPlayDelay:
@@ -997,6 +1024,7 @@
case S_Speak_utt: inqSpeakUtt(); break;
case S_Speak_len: inqSpeakLen(); break;
case S_Speak_stat: inqSpeakStat(); break;
+ case S_Speak_syncinterval: inqSpeakSyncinterval(); break;
case S_Log:
RepMsg( "rep Log = %s\n", slot_Log_file ); break;
case S_Log_conf:
@@ -1056,6 +1084,7 @@
case S_Speak_utt: prop_Speak_utt = prop; break;
case S_Speak_len: prop_Speak_len = prop; break;
case S_Speak_stat: prop_Speak_stat = prop; break;
+ case S_Speak_syncinterval: prop_Speak_syncinterval = prop; break;
default:
unknown_com();
}
Index: gtalk/read_conf.c
diff -u gtalk/read_conf.c:1.16 gtalk/read_conf.c:1.17
--- gtalk/read_conf.c:1.16 Sat Sep 2 09:58:09 2006
+++ gtalk/read_conf.c Thu Oct 19 12:26:33 2006
@@ -1,7 +1,8 @@
/* Copyright (c) 2000-2006 */
/* Yamashita Lab., Ritsumeikan University */
/* All rights reserved */
-
+/* */
+/* $Id: read_conf.c,v 1.17 2006/10/19 03:26:33 sako Exp $ */
#include <stdio.h>
#include <stdlib.h>
@@ -99,6 +100,8 @@
}
} else if( strcmp(cpara,"AUTO-PLAY-DELAY")==0 ) {
slot_Auto_play_delay = atoi( val );
+ } else if( strcmp(cpara,"SYNC-INTERVAL")==0 ) {
+ slot_Speak_syncinterval = atoi( val );
} else if( strcmp(cpara,"SPEAKER-ID")==0 ) {
if( n_speaker > 0 ) check_speaker_conf( n_speaker-1 );
speaker[n_speaker].code = malloc_char( val, "speaker.code" );
Index: gtalk/slot.h
diff -u gtalk/slot.h:1.14 gtalk/slot.h:1.15
--- gtalk/slot.h:1.14 Sat Sep 2 09:58:09 2006
+++ gtalk/slot.h Thu Oct 19 12:26:33 2006
@@ -1,6 +1,8 @@
/* Copyright (c) 2000-2006 */
/* Yamashita Lab., Ritsumeikan University */
/* All rights reserved */
+/* */
+/* $Id: slot.h,v 1.15 2006/10/19 03:26:33 sako Exp $ */
/* List of Slot */
@@ -22,6 +24,7 @@
#define S_Speak_utt 34
#define S_Speak_len 35
#define S_Speak_stat 36
+#define S_Speak_syncinterval 37
#define S_Save 40
#define S_SpeechFile 41
#define S_SavePros 42
@@ -67,6 +70,7 @@
{ S_Speak_utt, "Speak.utt" },
{ S_Speak_len, "Speak.len" },
{ S_Speak_stat, "Speak.stat" },
+ { S_Speak_syncinterval, "Speak.syncinterval" },
{ S_Save, "Save" },
{ S_SpeechFile, "SpeechFile" },
{ S_SavePros, "SavePros" },
@@ -113,6 +117,7 @@
extern SlotProp prop_Speak_utt;
extern SlotProp prop_Speak_len;
extern SlotProp prop_Speak_stat;
+extern SlotProp prop_Speak_syncinterval;
/* slots */
@@ -141,3 +146,4 @@
extern int slot_Log_aphrase;
extern int slot_Log_breath;
extern int slot_Log_sentence;
+extern int slot_Speak_syncinterval;