Sending refresh DTX packets every 400 ms independently of the encoded frame size.

Signed-off-by: Felicia Lim <flim@google.com>

1 files changed, 15 insertions, 11 deletions

diff --git a/src/opus_encoder.c b/src/opus_encoder.c
index 7b5f0abf..321bb2bb 100644
--- a/src/opus_encoder.c
+++ b/src/opus_encoder.c
@@ -112,7 +112,7 @@ struct OpusEncoder {
     opus_val16   delay_buffer[MAX_ENCODER_BUFFER*2];
 #ifndef DISABLE_FLOAT_API
     int          detected_bandwidth;
-    int          nb_no_activity_frames;
+    int          nb_no_activity_ms_Q1;
     opus_val32   peak_signal_energy;
 #endif
     int          nonfinal_frame; /* current frame is not the final in a packet */
@@ -893,24 +893,28 @@ static opus_val32 compute_frame_energy(const opus_val16 *pcm, int frame_size, in
 
 /* Decides if DTX should be turned on (=1) or off (=0) */
 static int decide_dtx_mode(opus_int activity,            /* indicates if this frame contains speech/music */
-                           int *nb_no_activity_frames    /* number of consecutive frames with no activity */
+                           int *nb_no_activity_ms_Q1,    /* number of consecutive milliseconds with no activity, in Q1 */
+                           int frame_size_ms_Q1          /* number of miliseconds in this update, in Q1 */
                            )
 
 {
    if (!activity)
    {
-      /* The number of consecutive DTX frames should be within the allowed bounds */
-      (*nb_no_activity_frames)++;
-      if (*nb_no_activity_frames > NB_SPEECH_FRAMES_BEFORE_DTX)
+      /* The number of consecutive DTX frames should be within the allowed bounds. 
+      Note that the allowed bound is defined in the Silk headers and assumes 20 ms
+      frames. As this function can be called with any frame length, a conversion to
+      miliseconds is done before the comparisons. */
+      (*nb_no_activity_ms_Q1) += frame_size_ms_Q1;
+      if (*nb_no_activity_ms_Q1 > NB_SPEECH_FRAMES_BEFORE_DTX*20*2)
       {
-         if (*nb_no_activity_frames <= (NB_SPEECH_FRAMES_BEFORE_DTX + MAX_CONSECUTIVE_DTX))
+         if (*nb_no_activity_ms_Q1 <= (NB_SPEECH_FRAMES_BEFORE_DTX + MAX_CONSECUTIVE_DTX)*20*2)
             /* Valid frame for DTX! */
             return 1;
          else
-            (*nb_no_activity_frames) = NB_SPEECH_FRAMES_BEFORE_DTX;
+            (*nb_no_activity_ms_Q1) = NB_SPEECH_FRAMES_BEFORE_DTX*20*2;
       }
    } else
-      (*nb_no_activity_frames) = 0;
+      (*nb_no_activity_ms_Q1) = 0;
 
    return 0;
 }
@@ -2132,7 +2136,7 @@ opus_int32 opus_encode_native(OpusEncoder *st, const opus_val16 *pcm, int frame_
 #ifndef DISABLE_FLOAT_API
     if (st->use_dtx && (analysis_info.valid || is_silence))
     {
-       if (decide_dtx_mode(activity, &st->nb_no_activity_frames))
+       if (decide_dtx_mode(activity, &st->nb_no_activity_ms_Q1, 2*1000*frame_size/st->Fs))
        {
           st->rangeFinal = 0;
           data[0] = gen_toc(st->mode, st->Fs/frame_size, curr_bandwidth, st->stream_channels);
@@ -2140,7 +2144,7 @@ opus_int32 opus_encode_native(OpusEncoder *st, const opus_val16 *pcm, int frame_
           return 1;
        }
     } else {
-       st->nb_no_activity_frames = 0;
+       st->nb_no_activity_ms_Q1 = 0;
     }
 #endif
 
@@ -2733,7 +2737,7 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
 #ifndef DISABLE_FLOAT_API
             else if (st->use_dtx) {
                 /* DTX determined by Opus. */
-                *value = st->nb_no_activity_frames >= NB_SPEECH_FRAMES_BEFORE_DTX;
+                *value = st->nb_no_activity_ms_Q1 >= NB_SPEECH_FRAMES_BEFORE_DTX*20*2;
             }
 #endif
             else {