tg2sip/libtgvoip/webrtc_dsp/api/audio/audio_frame.h

133 lines
4.8 KiB
C
Raw Normal View History

Squashed 'libtgvoip/' changes from 6053cf5..cfd62e6 cfd62e6 Why did it change the OS X project 3a58a16 2.4.3 c4a48b3 Updated OS X project 564eada Fix #63 4f64e2e fixes 0c732e2 fixes 12e76ed better logging f015b79 Merge pull request #62 from xvitaly/big-endian a1df90f Set preferred audio session parameters on iOS 59a975b Fixes 8fd89fc Fixes, mic level testing and volume adjustment 243acfa Backported WebRTC upstream patch with Big Endian support. fed3bb7 Detect when proxy does not support UDP and persist that across calls a7546d4 Merge commit '6d03dd9ae4bf48d7344341cdd2d055ebd3a6a42e' into public 6d03dd9 version 69adf70 Use server config for APM + iOS crash fix 0b42ec8 Update iOS project f1b9e63 packet logging beeea45 I apparently still suck at C++ memory management 24fceba Update project 7f54b91 crash fix f85ce99 Save more data in data saving mode f4c4f79 Collect packet stats and accept json string for server config 78e584c New protocol version: optimized packet size 8cf9177 Fixed build on iOS 9dd089d fixed build on android 5caaaaf Updated WebRTC APM cc0cf35 fixed deadlock 02f4835 Rearranged VoIPController methods and added sections 912f73d Updated OS X project 39376df Fixed audio glitches on Windows dfe1f03 Updated project 81daf3f fix 296187a Merge pull request #58 from telegramdesktop/tdesktop 44956ac Merge pull request #57 from UnigramDev/public fb0a2b0 Fix build for Linux. d6cf1b7 Updated UWP wrapper 0f06289 Merge branch 'public' of github.com:grishka/libtgvoip into public dcfad91 Fix #54 162f447 Merge pull request #56 from telegramdesktop/tdesktop a7ee511 Merge remote-tracking branch 'origin/tdesktop' into HEAD 467b148 Removed unused files b1a0b3d 2.3 9b292fd Fix warning in Xcode 10. 8d8522a Merge pull request #53 from UnigramDev/public 646f7d6 Merge branch 'public' into public 14d782b Fixes 68acf59 Added GetSignalBarsCount and GetConnectionState to CXWrapper 761c586 Added GetStats to CXWrapper f643b02 Prevent crash if UWP WASAPI devices aren't found b2ac10e Fixed UWP project 9a1ec51 Fixed build for Windows Phone, fixed some warnings 4aea54f fix git-subtree-dir: libtgvoip git-subtree-split: cfd62e66a825348ac51f49e5d20bf8827fef7a38
2019-02-06 18:22:38 +00:00
/*
* Copyright (c) 2018 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.
*/
#ifndef API_AUDIO_AUDIO_FRAME_H_
#define API_AUDIO_AUDIO_FRAME_H_
#include <stddef.h>
#include <stdint.h>
#include "rtc_base/constructormagic.h"
namespace webrtc {
/* This class holds up to 120 ms of super-wideband (32 kHz) stereo audio. It
* allows for adding and subtracting frames while keeping track of the resulting
* states.
*
* Notes
* - This is a de-facto api, not designed for external use. The AudioFrame class
* is in need of overhaul or even replacement, and anyone depending on it
* should be prepared for that.
* - The total number of samples is samples_per_channel_ * num_channels_.
* - Stereo data is interleaved starting with the left channel.
*/
class AudioFrame {
public:
// Using constexpr here causes linker errors unless the variable also has an
// out-of-class definition, which is impractical in this header-only class.
// (This makes no sense because it compiles as an enum value, which we most
// certainly cannot take the address of, just fine.) C++17 introduces inline
// variables which should allow us to switch to constexpr and keep this a
// header-only class.
enum : size_t {
// Stereo, 32 kHz, 120 ms (2 * 32 * 120)
// Stereo, 192 kHz, 20 ms (2 * 192 * 20)
kMaxDataSizeSamples = 7680,
kMaxDataSizeBytes = kMaxDataSizeSamples * sizeof(int16_t),
};
enum VADActivity { kVadActive = 0, kVadPassive = 1, kVadUnknown = 2 };
enum SpeechType {
kNormalSpeech = 0,
kPLC = 1,
kCNG = 2,
kPLCCNG = 3,
kUndefined = 4
};
AudioFrame();
// Resets all members to their default state.
void Reset();
// Same as Reset(), but leaves mute state unchanged. Muting a frame requires
// the buffer to be zeroed on the next call to mutable_data(). Callers
// intending to write to the buffer immediately after Reset() can instead use
// ResetWithoutMuting() to skip this wasteful zeroing.
void ResetWithoutMuting();
void UpdateFrame(uint32_t timestamp,
const int16_t* data,
size_t samples_per_channel,
int sample_rate_hz,
SpeechType speech_type,
VADActivity vad_activity,
size_t num_channels = 1);
void CopyFrom(const AudioFrame& src);
// Sets a wall-time clock timestamp in milliseconds to be used for profiling
// of time between two points in the audio chain.
// Example:
// t0: UpdateProfileTimeStamp()
// t1: ElapsedProfileTimeMs() => t1 - t0 [msec]
void UpdateProfileTimeStamp();
// Returns the time difference between now and when UpdateProfileTimeStamp()
// was last called. Returns -1 if UpdateProfileTimeStamp() has not yet been
// called.
int64_t ElapsedProfileTimeMs() const;
// data() returns a zeroed static buffer if the frame is muted.
// mutable_frame() always returns a non-static buffer; the first call to
// mutable_frame() zeros the non-static buffer and marks the frame unmuted.
const int16_t* data() const;
int16_t* mutable_data();
// Prefer to mute frames using AudioFrameOperations::Mute.
void Mute();
// Frame is muted by default.
bool muted() const;
// RTP timestamp of the first sample in the AudioFrame.
uint32_t timestamp_ = 0;
// Time since the first frame in milliseconds.
// -1 represents an uninitialized value.
int64_t elapsed_time_ms_ = -1;
// NTP time of the estimated capture time in local timebase in milliseconds.
// -1 represents an uninitialized value.
int64_t ntp_time_ms_ = -1;
size_t samples_per_channel_ = 0;
int sample_rate_hz_ = 0;
size_t num_channels_ = 0;
SpeechType speech_type_ = kUndefined;
VADActivity vad_activity_ = kVadUnknown;
// Monotonically increasing timestamp intended for profiling of audio frames.
// Typically used for measuring elapsed time between two different points in
// the audio path. No lock is used to save resources and we are thread safe
// by design. Also, absl::optional is not used since it will cause a "complex
// class/struct needs an explicit out-of-line destructor" build error.
int64_t profile_timestamp_ms_ = 0;
private:
// A permamently zeroed out buffer to represent muted frames. This is a
// header-only class, so the only way to avoid creating a separate empty
// buffer per translation unit is to wrap a static in an inline function.
static const int16_t* empty_data();
int16_t data_[kMaxDataSizeSamples];
bool muted_ = true;
RTC_DISALLOW_COPY_AND_ASSIGN(AudioFrame);
};
} // namespace webrtc
#endif // API_AUDIO_AUDIO_FRAME_H_