1 /*
2 * Copyright (c) 2014 The WebRTC project authors. All Rights Reserved.
3 *
4 * Use of this source code is governed by a BSD-style license
5 * that can be found in the LICENSE file in the root of the source
6 * tree. An additional intellectual property rights grant can be found
7 * in the file PATENTS. All contributing project authors may
8 * be found in the AUTHORS file in the root of the source tree.
9 */
10
11 #include "modules/audio_coding/codecs/g722/audio_encoder_g722.h"
12
13 #include <cstdint>
14
15 #include "modules/audio_coding/codecs/g722/g722_interface.h"
16 #include "rtc_base/checks.h"
17 #include "rtc_base/numerics/safe_conversions.h"
18
19 namespace webrtc {
20
21 namespace {
22
23 const size_t kSampleRateHz = 16000;
24
25 } // namespace
26
AudioEncoderG722Impl(const AudioEncoderG722Config & config,int payload_type)27 AudioEncoderG722Impl::AudioEncoderG722Impl(const AudioEncoderG722Config& config,
28 int payload_type)
29 : num_channels_(config.num_channels),
30 payload_type_(payload_type),
31 num_10ms_frames_per_packet_(
32 static_cast<size_t>(config.frame_size_ms / 10)),
33 num_10ms_frames_buffered_(0),
34 first_timestamp_in_buffer_(0),
35 encoders_(new EncoderState[num_channels_]),
36 interleave_buffer_(2 * num_channels_) {
37 RTC_CHECK(config.IsOk());
38 const size_t samples_per_channel =
39 kSampleRateHz / 100 * num_10ms_frames_per_packet_;
40 for (size_t i = 0; i < num_channels_; ++i) {
41 encoders_[i].speech_buffer.reset(new int16_t[samples_per_channel]);
42 encoders_[i].encoded_buffer.SetSize(samples_per_channel / 2);
43 }
44 Reset();
45 }
46
47 AudioEncoderG722Impl::~AudioEncoderG722Impl() = default;
48
SampleRateHz() const49 int AudioEncoderG722Impl::SampleRateHz() const {
50 return kSampleRateHz;
51 }
52
NumChannels() const53 size_t AudioEncoderG722Impl::NumChannels() const {
54 return num_channels_;
55 }
56
RtpTimestampRateHz() const57 int AudioEncoderG722Impl::RtpTimestampRateHz() const {
58 // The RTP timestamp rate for G.722 is 8000 Hz, even though it is a 16 kHz
59 // codec.
60 return kSampleRateHz / 2;
61 }
62
Num10MsFramesInNextPacket() const63 size_t AudioEncoderG722Impl::Num10MsFramesInNextPacket() const {
64 return num_10ms_frames_per_packet_;
65 }
66
Max10MsFramesInAPacket() const67 size_t AudioEncoderG722Impl::Max10MsFramesInAPacket() const {
68 return num_10ms_frames_per_packet_;
69 }
70
GetTargetBitrate() const71 int AudioEncoderG722Impl::GetTargetBitrate() const {
72 // 4 bits/sample, 16000 samples/s/channel.
73 return static_cast<int>(64000 * NumChannels());
74 }
75
Reset()76 void AudioEncoderG722Impl::Reset() {
77 num_10ms_frames_buffered_ = 0;
78 for (size_t i = 0; i < num_channels_; ++i)
79 RTC_CHECK_EQ(0, WebRtcG722_EncoderInit(encoders_[i].encoder));
80 }
81
82 absl::optional<std::pair<TimeDelta, TimeDelta>>
GetFrameLengthRange() const83 AudioEncoderG722Impl::GetFrameLengthRange() const {
84 return {{TimeDelta::Millis(num_10ms_frames_per_packet_ * 10),
85 TimeDelta::Millis(num_10ms_frames_per_packet_ * 10)}};
86 }
87
EncodeImpl(uint32_t rtp_timestamp,rtc::ArrayView<const int16_t> audio,rtc::Buffer * encoded)88 AudioEncoder::EncodedInfo AudioEncoderG722Impl::EncodeImpl(
89 uint32_t rtp_timestamp,
90 rtc::ArrayView<const int16_t> audio,
91 rtc::Buffer* encoded) {
92 if (num_10ms_frames_buffered_ == 0)
93 first_timestamp_in_buffer_ = rtp_timestamp;
94
95 // Deinterleave samples and save them in each channel's buffer.
96 const size_t start = kSampleRateHz / 100 * num_10ms_frames_buffered_;
97 for (size_t i = 0; i < kSampleRateHz / 100; ++i)
98 for (size_t j = 0; j < num_channels_; ++j)
99 encoders_[j].speech_buffer[start + i] = audio[i * num_channels_ + j];
100
101 // If we don't yet have enough samples for a packet, we're done for now.
102 if (++num_10ms_frames_buffered_ < num_10ms_frames_per_packet_) {
103 return EncodedInfo();
104 }
105
106 // Encode each channel separately.
107 RTC_CHECK_EQ(num_10ms_frames_buffered_, num_10ms_frames_per_packet_);
108 num_10ms_frames_buffered_ = 0;
109 const size_t samples_per_channel = SamplesPerChannel();
110 for (size_t i = 0; i < num_channels_; ++i) {
111 const size_t bytes_encoded = WebRtcG722_Encode(
112 encoders_[i].encoder, encoders_[i].speech_buffer.get(),
113 samples_per_channel, encoders_[i].encoded_buffer.data());
114 RTC_CHECK_EQ(bytes_encoded, samples_per_channel / 2);
115 }
116
117 const size_t bytes_to_encode = samples_per_channel / 2 * num_channels_;
118 EncodedInfo info;
119 info.encoded_bytes = encoded->AppendData(
120 bytes_to_encode, [&](rtc::ArrayView<uint8_t> encoded) {
121 // Interleave the encoded bytes of the different channels. Each separate
122 // channel and the interleaved stream encodes two samples per byte, most
123 // significant half first.
124 for (size_t i = 0; i < samples_per_channel / 2; ++i) {
125 for (size_t j = 0; j < num_channels_; ++j) {
126 uint8_t two_samples = encoders_[j].encoded_buffer.data()[i];
127 interleave_buffer_.data()[j] = two_samples >> 4;
128 interleave_buffer_.data()[num_channels_ + j] = two_samples & 0xf;
129 }
130 for (size_t j = 0; j < num_channels_; ++j)
131 encoded[i * num_channels_ + j] =
132 interleave_buffer_.data()[2 * j] << 4 |
133 interleave_buffer_.data()[2 * j + 1];
134 }
135
136 return bytes_to_encode;
137 });
138 info.encoded_timestamp = first_timestamp_in_buffer_;
139 info.payload_type = payload_type_;
140 info.encoder_type = CodecType::kG722;
141 return info;
142 }
143
EncoderState()144 AudioEncoderG722Impl::EncoderState::EncoderState() {
145 RTC_CHECK_EQ(0, WebRtcG722_CreateEncoder(&encoder));
146 }
147
~EncoderState()148 AudioEncoderG722Impl::EncoderState::~EncoderState() {
149 RTC_CHECK_EQ(0, WebRtcG722_FreeEncoder(encoder));
150 }
151
SamplesPerChannel() const152 size_t AudioEncoderG722Impl::SamplesPerChannel() const {
153 return kSampleRateHz / 100 * num_10ms_frames_per_packet_;
154 }
155
156 } // namespace webrtc
157