Merge branch 'master' of https://github.com/mannol1/toxcore

toxav/Makefile.inc

@@ -10,8 +10,8 @@ libtoxav_la_SOURCES = ../toxav/event.h \
                       ../toxav/rtp.c \
                       ../toxav/msi.h \
                       ../toxav/msi.c \
-                      ../toxav/media.h \
+                      ../toxav/codec.h \
-                      ../toxav/media.c \
+                      ../toxav/codec.c \
                       ../toxav/toxav.h \
                       ../toxav/toxav.c
 

toxav/codec.c

@@ -34,7 +34,10 @@
 #include <assert.h>
 
 #include "rtp.h"
-#include "media.h"
+#include "codec.h"
+
+const uint16_t min_jbuf_size = 10;
+const uint16_t min_readiness_idx = 6; /* when is buffer ready to dqq */
 
 int empty_queue(JitterBuffer *q)
 {
@@ -65,7 +68,7 @@ JitterBuffer *create_queue(int capacity)
     }
 
     q->size = 0;
-    q->capacity = capacity;
+    q->capacity = capacity >= min_jbuf_size ? capacity : min_jbuf_size;
     q->front = 0;
     q->rear = -1;
     q->queue_ready = 0;
@@ -141,14 +144,12 @@ void queue(JitterBuffer *q, RTPMessage *pk)
         empty_queue(q);
     }
 
-    if (q->size > 8)
+    if (q->size >= min_readiness_idx) q->queue_ready = 1;
-        q->queue_ready = 1;
 
     ++q->size;
     ++q->rear;
 
-    if (q->rear == q->capacity)
+    if (q->rear == q->capacity) q->rear = 0;
-        q->rear = 0;
 
     q->queue[q->rear] = pk;
 
@@ -175,8 +176,7 @@ void queue(JitterBuffer *q, RTPMessage *pk)
 
         a -= 1;
 
-        if (a < 0)
+        if (a < 0) a += q->capacity;
-            a += q->capacity;
     }
 }
 
@@ -264,6 +264,7 @@ CodecState *codec_init_session ( uint32_t audio_bitrate,
                                  uint16_t audio_frame_duration,
                                  uint32_t audio_sample_rate,
                                  uint32_t audio_channels,
+                                 uint32_t audio_VAD_tolerance_ms,
                                  uint16_t video_width,
                                  uint16_t video_height,
                                  uint32_t video_bitrate )
@@ -292,6 +293,10 @@ CodecState *codec_init_session ( uint32_t audio_bitrate,
         return NULL;
     }
 
+
+    retu->EVAD_tolerance = audio_VAD_tolerance_ms > audio_frame_duration ?
+                           audio_VAD_tolerance_ms / audio_frame_duration : audio_frame_duration;
+
     return retu;
 }
 
@@ -303,13 +308,36 @@ void codec_terminate_session ( CodecState *cs )
     if ( cs->audio_decoder )
         opus_decoder_destroy(cs->audio_decoder);
 
-
-    /* TODO: Terminate video
-     *           Do what?
-     */
     if ( cs->capabilities & v_decoding )
         vpx_codec_destroy(&cs->v_decoder);
 
     if ( cs->capabilities & v_encoding )
         vpx_codec_destroy(&cs->v_encoder);
 }
+
+inline float calculate_sum_sq (int16_t *n, uint16_t k)
+{
+    float result = 0;
+    uint16_t i = 0;
+
+    for ( ; i < k; i ++) result += (float) (n[i] * n[i]);
+
+    return result;
+}
+
+int energy_VAD(CodecState *cs, int16_t *PCM, uint16_t frame_size, float energy)
+{
+    float frame_energy = sqrt(calculate_sum_sq(PCM, frame_size)) / frame_size;
+
+    if ( frame_energy > energy) {
+        cs->EVAD_tolerance_cr = cs->EVAD_tolerance; /* Reset counter */
+        return 1;
+    }
+
+    if ( cs->EVAD_tolerance_cr ) {
+        cs->EVAD_tolerance_cr --;
+        return 1;
+    }
+
+    return 0;
+}

toxav/codec.h

@@ -21,8 +21,8 @@
  *
  */
 
-#ifndef _AVCODEC_H_
+#ifndef _CODEC_H_
-#define _AVCODEC_H_
+#define _CODEC_H_
 
 #include <stdio.h>
 #include <math.h>
@@ -46,6 +46,8 @@ typedef enum _Capabilities {
     v_decoding = 1 << 3
 } Capabilities;
 
+extern const uint16_t min_jbuf_size;
+
 typedef struct _CodecState {
 
     /* video encoding */
@@ -65,6 +67,9 @@ typedef struct _CodecState {
 
     uint64_t capabilities; /* supports*/
 
+    /* Voice activity detection */
+    uint32_t EVAD_tolerance; /* In frames */
+    uint32_t EVAD_tolerance_cr;
 } CodecState;
 
 
@@ -90,10 +95,15 @@ CodecState *codec_init_session ( uint32_t audio_bitrate,
                                  uint16_t audio_frame_duration,
                                  uint32_t audio_sample_rate,
                                  uint32_t audio_channels,
+                                 uint32_t audio_VAD_tolerance_ms,
                                  uint16_t video_width,
                                  uint16_t video_height,
-                                 uint32_t video_bitrate );
+                                 uint32_t video_bitrate);
 
 void codec_terminate_session(CodecState *cs);
 
-#endif
+
+/* Calculate energy and return 1 if has voice, 0 if not */
+int energy_VAD(CodecState *cs, int16_t *PCM, uint16_t frame_size, float energy);
+
+#endif /* _CODEC_H_ */

toxav/msi.c

@@ -257,7 +257,7 @@ iterator = iterator + 2 + _value_size; /* set iterator at new header or end_byte
             _it += 3; /* place it at the field value beginning */
             size_max -= 3;
 
-            switch ( _size ) { /* Compare the size of the hardcoded values ( vary fast and convenient ) */
+            switch ( _size ) { /* Compare the size of the hardcoded values ( very convenient ) */
 
                 case 4: { /* INFO header */
                     if ON_HEADER ( _it, size_max, msg->info, INFO_FIELD, 4 )

toxav/toxav.c

@@ -27,13 +27,12 @@
 #define _GNU_SOURCE /* implicit declaration warning */
 
 #include "rtp.h"
-#include "media.h"
+#include "codec.h"
 #include "msi.h"
 #include "toxav.h"
 
 #include "../toxcore/logger.h"
 
-
 #include <assert.h>
 #include <stdlib.h>
 #include <string.h>
@@ -78,7 +77,9 @@ const ToxAvCodecSettings av_DefaultSettings = {
     20,
     48000,
     1,
-    20
+    600,
+
+    10
 };
 
 
@@ -331,6 +332,7 @@ int toxav_prepare_transmission ( ToxAv *av, int32_t call_index, ToxAvCodecSettin
                                   codec_settings->audio_frame_duration,
                                   codec_settings->audio_sample_rate,
                                   codec_settings->audio_channels,
+                                  codec_settings->audio_VAD_tolerance,
                                   codec_settings->video_width,
                                   codec_settings->video_height,
                                   codec_settings->video_bitrate);
@@ -737,3 +739,10 @@ inline__ Tox *toxav_get_tox(ToxAv *av)
 {
     return (Tox *)av->messenger;
 }
+
+int toxav_has_activity(ToxAv *av, int32_t call_index, int16_t *PCM, uint16_t frame_size, float ref_energy)
+{
+    if ( !av->calls[call_index].cs ) return ErrorInvalidCodecState;
+
+    return energy_VAD(av->calls[call_index].cs, PCM, frame_size, ref_energy);
+}

toxav/toxav.h

@@ -90,6 +90,7 @@ typedef enum {
     ErrorTerminatingAudioRtp = -9, /* Returned in toxav_kill_transmission() */
     ErrorTerminatingVideoRtp = -10, /* Returned in toxav_kill_transmission() */
     ErrorPacketTooLarge = -11, /* Buffer exceeds size while encoding */
+    ErrorInvalidCodecState = -12, /* Codec state not initialized */
 
 } ToxAvError;
 
@@ -117,6 +118,7 @@ typedef struct _ToxAvCodecSettings {
     uint16_t audio_frame_duration; /* In ms */
     uint32_t audio_sample_rate; /* In Hz */
     uint32_t audio_channels;
+    uint32_t audio_VAD_tolerance; /* In ms */
 
     uint32_t jbuf_capacity; /* Size of jitter buffer */
 } ToxAvCodecSettings;
@@ -373,6 +375,8 @@ int toxav_set_video_queue_limit ( ToxAv *av, int32_t call_index, uint64_t limit
 
 Tox *toxav_get_tox(ToxAv *av);
 
+int toxav_has_activity ( ToxAv *av, int32_t call_index, int16_t *PCM, uint16_t frame_size, float ref_energy );
+
 #ifdef __cplusplus
 }
 #endif