← Back to branch summary

~ubuntu-branches/ubuntu/saucy/kopete/saucy-proposed

« back to all changes in this revision

Viewing changes to protocols/jabber/googletalk/libjingle/talk/session/phone/currentspeakermonitor.cc

  • Committer: Package Import Robot
  • Author(s): Jonathan Riddell
  • Date: 2013-06-21 02:22:39 UTC
  • Revision ID: package-import@ubuntu.com-20130621022239-63l3zc8p0nf26pt6
Tags: upstream-4.10.80
Import upstream version 4.10.80

Show diffs side-by-side

added added

removed removed

Lines of Context:
protocols/jabber/googletalk/libjingle/talk/session/phone/currentspeakermonitor.cc
 
1
/*
 
2
 * libjingle
 
3
 * Copyright 2011, Google Inc.
 
4
 *
 
5
 * Redistribution and use in source and binary forms, with or without
 
6
 * modification, are permitted provided that the following conditions are met:
 
7
 *
 
8
 *  1. Redistributions of source code must retain the above copyright notice,
 
9
 *     this list of conditions and the following disclaimer.
 
10
 *  2. Redistributions in binary form must reproduce the above copyright notice,
 
11
 *     this list of conditions and the following disclaimer in the documentation
 
12
 *     and/or other materials provided with the distribution.
 
13
 *  3. The name of the author may not be used to endorse or promote products
 
14
 *     derived from this software without specific prior written permission.
 
15
 *
 
16
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED
 
17
 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
 
18
 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO
 
19
 * EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 
20
 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
 
21
 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
 
22
 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
 
23
 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
 
24
 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
 
25
 * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
26
 */
 
27
 
 
28
#include "talk/session/phone/currentspeakermonitor.h"
 
29
 
 
30
#include "talk/base/logging.h"
 
31
#include "talk/session/phone/call.h"
 
32
 
 
33
namespace cricket {
 
34
 
 
35
namespace {
 
36
const int kMaxAudioLevel = 9;
 
37
// To avoid overswitching, we disable switching for a period of time after a
 
38
// switch is done.
 
39
const int kDefaultMinTimeBetweenSwitches = 1000;
 
40
}
 
41
 
 
42
CurrentSpeakerMonitor::CurrentSpeakerMonitor(Call* call, BaseSession* session)
 
43
    : started_(false),
 
44
      call_(call),
 
45
      session_(session),
 
46
      current_speaker_ssrc_(0),
 
47
      earliest_permitted_switch_time_(0),
 
48
      min_time_between_switches_(kDefaultMinTimeBetweenSwitches) {
 
49
}
 
50
 
 
51
CurrentSpeakerMonitor::~CurrentSpeakerMonitor() {
 
52
  Stop();
 
53
}
 
54
 
 
55
void CurrentSpeakerMonitor::Start() {
 
56
  if (!started_) {
 
57
    call_->SignalAudioMonitor.connect(
 
58
        this, &CurrentSpeakerMonitor::OnAudioMonitor);
 
59
    call_->SignalMediaStreamsUpdate.connect(
 
60
        this, &CurrentSpeakerMonitor::OnMediaStreamsUpdate);
 
61
 
 
62
    started_ = true;
 
63
  }
 
64
}
 
65
 
 
66
void CurrentSpeakerMonitor::Stop() {
 
67
  if (started_) {
 
68
    call_->SignalAudioMonitor.disconnect(this);
 
69
    call_->SignalMediaStreamsUpdate.disconnect(this);
 
70
 
 
71
    started_ = false;
 
72
    ssrc_to_speaking_state_map_.clear();
 
73
    current_speaker_ssrc_ = 0;
 
74
    earliest_permitted_switch_time_ = 0;
 
75
  }
 
76
}
 
77
 
 
78
void CurrentSpeakerMonitor::set_min_time_between_switches(
 
79
    uint32 min_time_between_switches) {
 
80
  min_time_between_switches_ = min_time_between_switches;
 
81
}
 
82
 
 
83
void CurrentSpeakerMonitor::OnAudioMonitor(Call* call, const AudioInfo& info) {
 
84
  std::map<uint32, int> active_ssrc_to_level_map;
 
85
  cricket::AudioInfo::StreamList::const_iterator stream_list_it;
 
86
  for (stream_list_it = info.active_streams.begin();
 
87
       stream_list_it != info.active_streams.end(); ++stream_list_it) {
 
88
    uint32 ssrc = stream_list_it->first;
 
89
    active_ssrc_to_level_map[ssrc] = stream_list_it->second;
 
90
 
 
91
    // It's possible we haven't yet added this source to our map.  If so,
 
92
    // add it now with a "not speaking" state.
 
93
    if (ssrc_to_speaking_state_map_.find(ssrc) ==
 
94
        ssrc_to_speaking_state_map_.end()) {
 
95
      ssrc_to_speaking_state_map_[ssrc] = SS_NOT_SPEAKING;
 
96
    }
 
97
  }
 
98
 
 
99
  int max_level = 0;
 
100
  uint32 loudest_speaker_ssrc = 0;
 
101
 
 
102
  // Update the speaking states of all participants based on the new audio
 
103
  // level information.  Also retain loudest speaker.
 
104
  std::map<uint32, SpeakingState>::iterator state_it;
 
105
  for (state_it = ssrc_to_speaking_state_map_.begin();
 
106
       state_it != ssrc_to_speaking_state_map_.end(); ++state_it) {
 
107
    bool is_previous_speaker = current_speaker_ssrc_ == state_it->first;
 
108
 
 
109
    // This uses a state machine in order to gradually identify
 
110
    // members as having started or stopped speaking. Matches the
 
111
    // algorithm used by the hangouts js code.
 
112
 
 
113
    std::map<uint32, int>::const_iterator level_it =
 
114
        active_ssrc_to_level_map.find(state_it->first);
 
115
    // Note that the stream map only contains streams with non-zero audio
 
116
    // levels.
 
117
    int level = (level_it != active_ssrc_to_level_map.end()) ?
 
118
        level_it->second : 0;
 
119
    switch (state_it->second) {
 
120
      case SS_NOT_SPEAKING:
 
121
        if (level > 0) {
 
122
          // Reset level because we don't think they're really speaking.
 
123
          level = 0;
 
124
          state_it->second = SS_MIGHT_BE_SPEAKING;
 
125
        } else {
 
126
          // State unchanged.
 
127
        }
 
128
        break;
 
129
      case SS_MIGHT_BE_SPEAKING:
 
130
        if (level > 0) {
 
131
          state_it->second = SS_SPEAKING;
 
132
        } else {
 
133
          state_it->second = SS_NOT_SPEAKING;
 
134
        }
 
135
        break;
 
136
      case SS_SPEAKING:
 
137
        if (level > 0) {
 
138
          // State unchanged.
 
139
        } else {
 
140
          state_it->second = SS_WAS_SPEAKING_RECENTLY1;
 
141
          if (is_previous_speaker) {
 
142
            // Assume this is an inter-word silence and assign him the highest
 
143
            // volume.
 
144
            level = kMaxAudioLevel;
 
145
          }
 
146
        }
 
147
        break;
 
148
      case SS_WAS_SPEAKING_RECENTLY1:
 
149
        if (level > 0) {
 
150
          state_it->second = SS_SPEAKING;
 
151
        } else {
 
152
          state_it->second = SS_WAS_SPEAKING_RECENTLY2;
 
153
          if (is_previous_speaker) {
 
154
            // Assume this is an inter-word silence and assign him the highest
 
155
            // volume.
 
156
            level = kMaxAudioLevel;
 
157
          }
 
158
        }
 
159
        break;
 
160
      case SS_WAS_SPEAKING_RECENTLY2:
 
161
        if (level > 0) {
 
162
          state_it->second = SS_SPEAKING;
 
163
        } else {
 
164
          state_it->second = SS_NOT_SPEAKING;
 
165
        }
 
166
        break;
 
167
    }
 
168
 
 
169
    if (level > max_level) {
 
170
      loudest_speaker_ssrc = state_it->first;
 
171
      max_level = level;
 
172
    } else if (level > 0 && level == max_level && is_previous_speaker) {
 
173
      // Favor continuity of loudest speakers if audio levels are equal.
 
174
      loudest_speaker_ssrc = state_it->first;
 
175
    }
 
176
  }
 
177
 
 
178
  // We avoid over-switching by disabling switching for a period of time after
 
179
  // a switch is done.
 
180
  uint32 now = talk_base::Time();
 
181
  if (earliest_permitted_switch_time_ <= now &&
 
182
      current_speaker_ssrc_ != loudest_speaker_ssrc) {
 
183
    current_speaker_ssrc_ = loudest_speaker_ssrc;
 
184
    LOG(LS_INFO) << "Current speaker changed to " << current_speaker_ssrc_;
 
185
    earliest_permitted_switch_time_ = now + min_time_between_switches_;
 
186
    SignalUpdate(this, current_speaker_ssrc_);
 
187
  }
 
188
}
 
189
 
 
190
void CurrentSpeakerMonitor::OnMediaStreamsUpdate(Call* call,
 
191
                                                 Session* session,
 
192
                                                 const MediaStreams& added,
 
193
                                                 const MediaStreams& removed) {
 
194
  if (call == call_ && session == session_) {
 
195
    // Update the speaking state map based on added and removed streams.
 
196
    for (std::vector<cricket::StreamParams>::const_iterator
 
197
           it = removed.video().begin(); it != removed.video().end(); ++it) {
 
198
      ssrc_to_speaking_state_map_.erase(it->first_ssrc());
 
199
    }
 
200
 
 
201
    for (std::vector<cricket::StreamParams>::const_iterator
 
202
           it = added.video().begin(); it != added.video().end(); ++it) {
 
203
      ssrc_to_speaking_state_map_[it->first_ssrc()] = SS_NOT_SPEAKING;
 
204
    }
 
205
  }
 
206
}
 
207
 
 
208
}  // namespace cricket