| 1 |
/* Example of simple PocketSphinx speech segmentation. |
| 2 |
* |
| 3 |
* MIT license (c) 2022, see LICENSE for more information. |
| 4 |
* |
| 5 |
* Author: David Huggins-Daines <[email protected]> |
| 6 |
*/ |
| 7 |
/** |
| 8 |
* @example live_pulseaudio.c |
| 9 |
* @brief Speech recognition with live audio input and endpointing. |
| 10 |
* |
| 11 |
* This file shows how to use PocketSphinx with microphone input using |
| 12 |
* PulseAudio. |
| 13 |
* |
| 14 |
* To compile it, assuming you have built the library as in |
| 15 |
* \ref unix_install "these directions", you can run: |
| 16 |
* |
| 17 |
* cmake --build build --target live_pulseaudio |
| 18 |
* |
| 19 |
* Alternately, if PocketSphinx is installed system-wide, you can run: |
| 20 |
* |
| 21 |
* gcc -o live_pulseaudio live_pulseaudio.c \ |
| 22 |
* $(pkg-config --libs --cflags pocketsphinx libpulse-simple) |
| 23 |
* |
| 24 |
* |
| 25 |
*/ |
| 26 |
#include <pulse/simple.h> |
| 27 |
#include <pulse/error.h> |
| 28 |
#include <pocketsphinx.h> |
| 29 |
#include <signal.h> |
| 30 |
|
| 31 |
static int global_done = 0; |
| 32 |
static void |
| 33 |
catch_sig(int signum) |
| 34 |
{ |
| 35 |
(void)signum; |
| 36 |
global_done = 1; |
| 37 |
} |
| 38 |
|
| 39 |
int |
| 40 |
main(int argc, char *argv[]) |
| 41 |
{ |
| 42 |
|
| 43 |
pa_simple *s; |
| 44 |
pa_sample_spec ss; |
| 45 |
int err; |
| 46 |
ps_decoder_t *decoder; |
| 47 |
ps_config_t *config; |
| 48 |
ps_endpointer_t *ep; |
| 49 |
short *frame; |
| 50 |
size_t frame_size; |
| 51 |
|
| 52 |
(void)argc; (void)argv; |
| 53 |
|
| 54 |
config = ps_config_init(NULL); |
| 55 |
ps_default_search_args(config); |
| 56 |
if ((decoder = ps_init(config)) == NULL) |
| 57 |
E_FATAL("PocketSphinx decoder init failed\n"); |
| 58 |
if ((ep = ps_endpointer_init(0, 0.0, 0, 0, 0)) == NULL) |
| 59 |
E_FATAL("PocketSphinx endpointer init failed\n"); |
| 60 |
frame_size = ps_endpointer_frame_size(ep); |
| 61 |
if ((frame = malloc(frame_size * sizeof(frame[0]))) == NULL) |
| 62 |
E_FATAL_SYSTEM("Failed to allocate frame"); |
| 63 |
|
| 64 |
ss.format = PA_SAMPLE_S16NE; |
| 65 |
ss.channels = 1; |
| 66 |
ss.rate = ps_config_int(config, "samprate"); |
| 67 |
if ((s = pa_simple_new(NULL, "live_pulseaudio", PA_STREAM_RECORD, NULL, |
| 68 |
"live", &ss, NULL, NULL, &err)) == NULL) |
| 69 |
E_FATAL("Failed to connect to PulseAudio: %s\n", |
| 70 |
pa_strerror(err)); |
| 71 |
if (signal(SIGINT, catch_sig) == SIG_ERR) |
| 72 |
E_FATAL_SYSTEM("Failed to set SIGINT handler"); |
| 73 |
while (!global_done) { |
| 74 |
const int16 *speech; |
| 75 |
int prev_in_speech = ps_endpointer_in_speech(ep); |
| 76 |
if (pa_simple_read(s, frame, |
| 77 |
frame_size * sizeof(frame[0]), &err) < 0) { |
| 78 |
E_ERROR("Error in pa_simple_read: %s\n", |
| 79 |
pa_strerror(err)); |
| 80 |
break; |
| 81 |
} |
| 82 |
speech = ps_endpointer_process(ep, frame); |
| 83 |
if (speech != NULL) { |
| 84 |
const char *hyp; |
| 85 |
if (!prev_in_speech) { |
| 86 |
fprintf(stderr, "Speech start at %.2f\n", |
| 87 |
ps_endpointer_speech_start(ep)); |
| 88 |
ps_start_utt(decoder); |
| 89 |
} |
| 90 |
if (ps_process_raw(decoder, speech, frame_size, FALSE, FALSE) < 0) |
| 91 |
E_FATAL("ps_process_raw() failed\n"); |
| 92 |
if ((hyp = ps_get_hyp(decoder, NULL)) != NULL) |
| 93 |
fprintf(stderr, "PARTIAL RESULT: %s\n", hyp); |
| 94 |
if (!ps_endpointer_in_speech(ep)) { |
| 95 |
fprintf(stderr, "Speech end at %.2f\n", |
| 96 |
ps_endpointer_speech_end(ep)); |
| 97 |
ps_end_utt(decoder); |
| 98 |
if ((hyp = ps_get_hyp(decoder, NULL)) != NULL) |
| 99 |
printf("%s\n", hyp); |
| 100 |
} |
| 101 |
} |
| 102 |
} |
| 103 |
pa_simple_free(s); |
| 104 |
free(frame); |
| 105 |
ps_endpointer_free(ep); |
| 106 |
ps_free(decoder); |
| 107 |
ps_config_free(config); |
| 108 |
|
| 109 |
return 0; |
| 110 |
} |