started opus encoding implamantationsound_detector

2 files changed, 52 insertions, 23 deletions

diff --git a/sound_detector/Makefile b/sound_detector/Makefile
index c38b514..dc0049b 100644
--- a/sound_detector/Makefile
+++ b/sound_detector/Makefile
@@ -1,6 +1,6 @@
 all:
 	g++ -c main.cpp -std=gnu++11 -o main.o
-	g++ -o sound_detector main.o -lboost_system -lboost_date_time -lboost_thread -lboost_random-mt -lboost_filesystem -lportaudio -lvorbis -lvorbisenc -logg -Wl,-O1 -s
+	g++ -o sound_detector main.o -lboost_system -lboost_date_time -lboost_thread -lboost_random-mt -lboost_filesystem -lportaudio -lvorbis -lvorbisenc -logg -lopus -Wl,-O1 -s
 clean:
 	rm *.o
 
diff --git a/sound_detector/main.cpp b/sound_detector/main.cpp
index 6bffcb6..e212d7d 100644
--- a/sound_detector/main.cpp
+++ b/sound_detector/main.cpp
@@ -32,7 +32,6 @@
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
-//#include <complex.h>
 
 //c++
 #include <limits>
@@ -43,7 +42,12 @@
 //portaudio
 #include <portaudio.h>
 
+//vorbis
 #include <vorbis/vorbisenc.h>
+
+//opus
+#include <opus/opus.h>
+
 //boost
 #include <boost/thread.hpp>
 #include <boost/bind.hpp>
@@ -68,6 +72,14 @@ boost::mutex lock;
 
 std::vector<int16_t> buffer;
 
+struct detector_params
+{
+	std::string out_dir;
+	double sample_rate;
+};
+
+detector_params sd_params;
+
 struct vorbis_params
 {
 	ogg_stream_state os;
@@ -79,7 +91,7 @@ struct vorbis_params
 	vorbis_block vb;
 };
 
-void encode_start(vorbis_params &vparams)
+void vorbis_encode_start(vorbis_params &vparams)
 {
 	vorbis_info_init(&(vparams.vi));
 	if(vorbis_encode_init_vbr(&(vparams.vi), 1, rate,.0))
@@ -107,7 +119,7 @@ void encode_start(vorbis_params &vparams)
 	}
 }
 
-void encode_data(vorbis_params &vparams)
+void vorbis_encode_data(vorbis_params &vparams)
 {
 	float **b = vorbis_analysis_buffer(&(vparams.vd), buffer.size());
 	for(size_t i = 0; i < buffer.size(); i++)
@@ -136,7 +148,7 @@ void encode_data(vorbis_params &vparams)
 	}
 }
 
-void encode_end(vorbis_params &vparams)
+void vorbis_encode_end(vorbis_params &vparams)
 {
 	vorbis_analysis_wrote(&(vparams.vd),0);
 	ogg_stream_clear(&(vparams.os));
@@ -146,6 +158,23 @@ void encode_end(vorbis_params &vparams)
 	vorbis_info_clear(&(vparams.vi));
 }
 
+struct opus_params
+{
+	OpusEncoder *encoder;
+};
+
+void opus_encode_start(opus_params &op)
+{
+	if(op.encoder)
+		free(op.encoder);
+	int encoder_error = 0;
+	op.encoder = opus_encoder_create(sd_params.sample_rate, 1, OPUS_APPLICATION_AUDIO, &encoder_error);
+	opus_encoder_ctl(op.encoder, OPUS_SET_BITRATE(24000));
+	opus_encoder_ctl(op.encoder, OPUS_SET_COMPLEXITY(10));
+	opus_encoder_ctl(op.encoder, OPUS_SET_SIGNAL(OPUS_SIGNAL_MUSIC));
+
+}
+
 std::string time_str()
 {
 	boost::posix_time::ptime now = boost::posix_time::second_clock::local_time();
@@ -156,7 +185,6 @@ std::string time_str()
 #endif
 }
 
-std::string out_dir;
 
 
 void handle_data()
@@ -182,13 +210,13 @@ void handle_data()
 				{
 					if(debug)
 						printf("write started\n");
-					boost::filesystem::path p(out_dir);
+					boost::filesystem::path p(sd_params.out_dir);
 					p += "/";
 					p += time_str();
 					p += ".ogg";
 					file = fopen(p.string().c_str(), "wb");
-					encode_start(vparams);
-					encode_data(vparams);
+					vorbis_encode_start(vparams);
+					vorbis_encode_data(vparams);
 					sound_detected = true;
 				}
 				else
@@ -207,7 +235,7 @@ void handle_data()
 				for(size_t i = 0; i < buffer.size(); i++)
 					level += abs(buffer[i]);
 				level /= buffer.size();
-				encode_data(vparams);
+				vorbis_encode_data(vparams);
 				if(debug)
 					printf("level detected %f, level lower than %f required to stop\n", ((float)level/(float)INT16_MAX)*100, stop_thresold_percent);
 				lock.unlock();
@@ -216,7 +244,7 @@ void handle_data()
 					if(debug)
 						printf("write stopped\n");
 					fclose(file);
-					encode_end(vparams);
+					vorbis_encode_end(vparams);
 					sound_detected = false;
 				}
 			}
@@ -323,7 +351,7 @@ int main(int argc, char **argv)
 			stop_thresold_percent = strtof(optarg, NULL);
 			break;
 			case 'o':
-			out_dir = optarg;
+			sd_params.out_dir = optarg;
 			break;
 			case 'd':
 			device = atoi(optarg);
@@ -348,17 +376,17 @@ int main(int argc, char **argv)
 		printf("ERROR: device number is required\n");
 		return 1;
 	}
-	if(out_dir.empty())
+	if(sd_params.out_dir.empty())
 	{
 		printf("ERROR: output directory is required\n");
 		return 1;
 	}
-	if(!boost::filesystem::exists(out_dir))
+	if(!boost::filesystem::exists(sd_params.out_dir))
 	{
 		printf("ERROR: output directory does not exists\n");
 		return 1;
 	}
-	if(!boost::filesystem::is_directory(out_dir))
+	if(!boost::filesystem::is_directory(sd_params.out_dir))
 	{
 		printf("ERROR: output directory is not directory %%) \n");
 		return 1;
@@ -366,16 +394,17 @@ int main(int argc, char **argv)
 	thresold = (int)(((float)INT16_MAX/100.0)*thresold_percent);
 	stop_thresold = (int)(((float)INT16_MAX/100.0)*stop_thresold_percent);
 	PaStream *stream;
-	PaStreamParameters params;
-	memset(&params, 0, sizeof(PaStreamParameters));
-	params.channelCount = 1;
-	params.sampleFormat = paInt16 | paNonInterleaved;
-	params.device = device;
+	PaStreamParameters pa_params;
+	memset(&pa_params, 0, sizeof(PaStreamParameters));
+	pa_params.channelCount = 1;
+	pa_params.sampleFormat = paInt16 | paNonInterleaved;
+	pa_params.device = device;
 	const PaDeviceInfo *info = Pa_GetDeviceInfo(device);
-	params.suggestedLatency = info->defaultHighInputLatency;
+	pa_params.suggestedLatency = info->defaultHighInputLatency;
+	sd_params.sample_rate = get_lowest_rate(&pa_params);
 	if(debug)
-		printf("choosen device latency %f, choosen rate %f\n", info->defaultHighInputLatency, get_lowest_rate(&params));
-	err = Pa_OpenStream(&stream, &params, NULL, get_lowest_rate(&params), paFramesPerBufferUnspecified, paClipOff, stream_callback, NULL);
+		printf("choosen device latency %f, choosen rate %f\n", info->defaultHighInputLatency, sd_params.sample_rate);
+	err = Pa_OpenStream(&stream, &pa_params, NULL, sd_params.sample_rate, paFramesPerBufferUnspecified, paClipOff, stream_callback, NULL);
 	if(err != paNoError)
 		printf("PortAudio error: %s\n", Pa_GetErrorText(err));
 #ifndef WIN32