| /* |
| * Copyright (c) 2014 The WebRTC project authors. All Rights Reserved. |
| * |
| * Use of this source code is governed by a BSD-style license |
| * that can be found in the LICENSE file in the root of the source |
| * tree. An additional intellectual property rights grant can be found |
| * in the file PATENTS. All contributing project authors may |
| * be found in the AUTHORS file in the root of the source tree. |
| */ |
| |
| #include "webrtc/modules/audio_coding/codecs/g722/include/audio_encoder_g722.h" |
| |
| #include <limits> |
| #include "webrtc/base/checks.h" |
| #include "webrtc/modules/audio_coding/codecs/g722/include/g722_interface.h" |
| |
| namespace webrtc { |
| |
| namespace { |
| |
| const int kSampleRateHz = 16000; |
| |
| } // namespace |
| |
| AudioEncoderG722::EncoderState::EncoderState() { |
| CHECK_EQ(0, WebRtcG722_CreateEncoder(&encoder)); |
| CHECK_EQ(0, WebRtcG722_EncoderInit(encoder)); |
| } |
| |
| AudioEncoderG722::EncoderState::~EncoderState() { |
| CHECK_EQ(0, WebRtcG722_FreeEncoder(encoder)); |
| } |
| |
| AudioEncoderG722::AudioEncoderG722(const Config& config) |
| : num_channels_(config.num_channels), |
| payload_type_(config.payload_type), |
| num_10ms_frames_per_packet_(config.frame_size_ms / 10), |
| num_10ms_frames_buffered_(0), |
| first_timestamp_in_buffer_(0), |
| encoders_(new EncoderState[num_channels_]), |
| interleave_buffer_(new uint8_t[2 * num_channels_]) { |
| CHECK_EQ(config.frame_size_ms % 10, 0) |
| << "Frame size must be an integer multiple of 10 ms."; |
| const int samples_per_channel = |
| kSampleRateHz / 100 * num_10ms_frames_per_packet_; |
| for (int i = 0; i < num_channels_; ++i) { |
| encoders_[i].speech_buffer.reset(new int16_t[samples_per_channel]); |
| encoders_[i].encoded_buffer.reset(new uint8_t[samples_per_channel / 2]); |
| } |
| } |
| |
| AudioEncoderG722::~AudioEncoderG722() {} |
| |
| int AudioEncoderG722::SampleRateHz() const { |
| return kSampleRateHz; |
| } |
| |
| int AudioEncoderG722::RtpTimestampRateHz() const { |
| // The RTP timestamp rate for G.722 is 8000 Hz, even though it is a 16 kHz |
| // codec. |
| return kSampleRateHz / 2; |
| } |
| |
| int AudioEncoderG722::NumChannels() const { |
| return num_channels_; |
| } |
| |
| size_t AudioEncoderG722::MaxEncodedBytes() const { |
| return static_cast<size_t>(SamplesPerChannel() / 2 * num_channels_); |
| } |
| |
| int AudioEncoderG722::Num10MsFramesInNextPacket() const { |
| return num_10ms_frames_per_packet_; |
| } |
| |
| int AudioEncoderG722::Max10MsFramesInAPacket() const { |
| return num_10ms_frames_per_packet_; |
| } |
| |
| AudioEncoder::EncodedInfo AudioEncoderG722::EncodeInternal( |
| uint32_t rtp_timestamp, |
| const int16_t* audio, |
| size_t max_encoded_bytes, |
| uint8_t* encoded) { |
| CHECK_GE(max_encoded_bytes, MaxEncodedBytes()); |
| |
| if (num_10ms_frames_buffered_ == 0) |
| first_timestamp_in_buffer_ = rtp_timestamp; |
| |
| // Deinterleave samples and save them in each channel's buffer. |
| const int start = kSampleRateHz / 100 * num_10ms_frames_buffered_; |
| for (int i = 0; i < kSampleRateHz / 100; ++i) |
| for (int j = 0; j < num_channels_; ++j) |
| encoders_[j].speech_buffer[start + i] = audio[i * num_channels_ + j]; |
| |
| // If we don't yet have enough samples for a packet, we're done for now. |
| if (++num_10ms_frames_buffered_ < num_10ms_frames_per_packet_) { |
| return kZeroEncodedBytes; |
| } |
| |
| // Encode each channel separately. |
| CHECK_EQ(num_10ms_frames_buffered_, num_10ms_frames_per_packet_); |
| num_10ms_frames_buffered_ = 0; |
| const int samples_per_channel = SamplesPerChannel(); |
| for (int i = 0; i < num_channels_; ++i) { |
| const int encoded = WebRtcG722_Encode( |
| encoders_[i].encoder, encoders_[i].speech_buffer.get(), |
| samples_per_channel, encoders_[i].encoded_buffer.get()); |
| CHECK_GE(encoded, 0); |
| CHECK_EQ(encoded, samples_per_channel / 2); |
| } |
| |
| // Interleave the encoded bytes of the different channels. Each separate |
| // channel and the interleaved stream encodes two samples per byte, most |
| // significant half first. |
| for (int i = 0; i < samples_per_channel / 2; ++i) { |
| for (int j = 0; j < num_channels_; ++j) { |
| uint8_t two_samples = encoders_[j].encoded_buffer[i]; |
| interleave_buffer_[j] = two_samples >> 4; |
| interleave_buffer_[num_channels_ + j] = two_samples & 0xf; |
| } |
| for (int j = 0; j < num_channels_; ++j) |
| encoded[i * num_channels_ + j] = |
| interleave_buffer_[2 * j] << 4 | interleave_buffer_[2 * j + 1]; |
| } |
| EncodedInfo info; |
| info.encoded_bytes = samples_per_channel / 2 * num_channels_; |
| info.encoded_timestamp = first_timestamp_in_buffer_; |
| info.payload_type = payload_type_; |
| return info; |
| } |
| |
| int AudioEncoderG722::SamplesPerChannel() const { |
| return kSampleRateHz / 100 * num_10ms_frames_per_packet_; |
| } |
| |
| } // namespace webrtc |