FaceAccess/Record/RkAudio.cpp
2024-09-04 17:57:23 +08:00

265 lines
8.2 KiB
C++

#include "RkAudio.h"
#include "BoostLog.h"
#include <cstring>
#include <rkmedia/rkmedia_api.h>
namespace RkAudio {
constexpr RK_U32 VqeFrameSample = 256; // 16ms;
constexpr auto AudioNode = "hw:0,0";
constexpr auto ParamFilePath = "/system/etc/RKAP_3A_Para.bin";
static SAMPLE_FORMAT_E rkAiFormat(Format::SampleType sampleType) {
SAMPLE_FORMAT_E ret = RK_SAMPLE_FMT_NONE;
switch (sampleType) {
case Format::SampleType::Unknown:
ret = RK_SAMPLE_FMT_NONE;
break;
case Format::SampleType::SignedInt16:
ret = RK_SAMPLE_FMT_S16;
break;
case Format::SampleType::SignedInt:
ret = RK_SAMPLE_FMT_S32;
break;
case Format::SampleType::Float:
ret = RK_SAMPLE_FMT_FLT;
break;
default:
LOG(error) << "unkonwn sample type: " << static_cast<int>(sampleType);
ret = RK_SAMPLE_FMT_NONE;
break;
}
return ret;
}
Input::Input() {
}
Input::~Input() {
if (m_channel >= 0) {
stop();
}
}
bool Input::open(const Format &format, bool enableVqe) {
bool ret = false;
m_channel = 0;
AI_CHN_ATTR_S parameter = {0};
parameter.pcAudioNode = (RK_CHAR *)AudioNode;
parameter.enAiLayout = AI_LAYOUT_MIC_REF; // remove ref channel, and output mic mono
parameter.enSampleFormat = rkAiFormat(format.sampleType);
parameter.u32Channels = format.channels;
parameter.u32SampleRate = format.sampleRate;
parameter.u32NbSamples = format.sampleRate / 1000 * format.period;
int status = RK_MPI_AI_SetChnAttr(m_channel, &parameter);
if (status) {
LOG(error) << "RK_MPI_AI_SetChnAttr() failed, status: " << status;
return ret;
}
status = RK_MPI_AI_EnableChn(m_channel);
if (status) {
LOG(error) << "RK_MPI_AI_EnableChn() failed, status: " << status;
return ret;
}
if (enableVqe) {
AI_TALKVQE_CONFIG_S config = {0};
status = RK_MPI_AI_GetTalkVqeAttr(m_channel, &config);
if (status) {
LOG(error) << "RK_MPI_AI_GetTalkVqeAttr() failed, status: " << status;
return ret;
}
LOG(info) << "param file: " << config.aParamFilePath;
config.s32WorkSampleRate = format.sampleRate;
config.s32FrameSample = VqeFrameSample;
config.u32OpenMask = AI_TALKVQE_MASK_AEC | AI_TALKVQE_MASK_ANR | AI_TALKVQE_MASK_AGC;
strncpy(config.aParamFilePath, ParamFilePath, sizeof(config.aParamFilePath));
RK_MPI_AI_SetTalkVqeAttr(m_channel, &config);
if (status) {
LOG(error) << "RK_MPI_AI_SetTalkVqeAttr() failed, status: " << status;
return ret;
}
status = RK_MPI_AI_EnableVqe(m_channel);
if (status) {
LOG(error) << "RK_MPI_AI_EnableVqe() failed, status: " << status;
return ret;
}
}
status = RK_MPI_AI_StartStream(0);
if (status) {
LOG(info) << "start AI failed, status: " << status;
return ret;
}
m_exit = false;
m_thread = std::thread(&Input::run, this);
ret = true;
return ret;
}
void Input::stop() {
m_exit = true;
if (m_thread.joinable()) m_thread.join();
if (m_channel >= 0) {
RK_MPI_AI_DisableVqe(m_channel);
RK_MPI_AI_DisableChn(m_channel);
m_channel = -1;
}
}
void Input::setDataCallback(const ReadCallback &callback) {
m_callback = callback;
}
void Input::run() {
while (!m_exit) {
auto mediaBuffer = RK_MPI_SYS_GetMediaBuffer(RK_ID_AI, 0, -1);
if (!mediaBuffer) {
LOG(error) << "RK_MPI_SYS_GetMediaBuffer() failed.";
continue;
}
if (m_callback) {
Frame frame;
frame.data = reinterpret_cast<uint8_t *>(RK_MPI_MB_GetPtr(mediaBuffer));
frame.byteSize = RK_MPI_MB_GetSize(mediaBuffer);
frame.frameSize = frame.byteSize / m_format.channels / sizeof(uint16_t);
frame.timestamp = std::chrono::system_clock::now();
m_callback(frame);
}
RK_MPI_MB_ReleaseBuffer(mediaBuffer);
}
}
Output::Output() {
}
Output::~Output() {
close();
}
bool Output::open(uint32_t sampleSize, uint32_t sampleRate, uint32_t channels, uint32_t period, bool enableVqe) {
m_channel = 0;
AO_CHN_ATTR_S parameter = {0};
parameter.pcAudioNode = (RK_CHAR *)AudioNode;
parameter.enSampleFormat = RK_SAMPLE_FMT_S16;
parameter.u32NbSamples = sampleRate / 1000 * period;
parameter.u32SampleRate = sampleRate;
parameter.u32Channels = channels;
RK_MPI_AO_SetChnAttr(m_channel, &parameter);
RK_MPI_AO_EnableChn(m_channel);
if (enableVqe) {
AO_VQE_CONFIG_S config = {0};
config.s32WorkSampleRate = sampleRate;
config.s32FrameSample = VqeFrameSample;
config.u32OpenMask = AO_VQE_MASK_ANR | AO_VQE_MASK_AGC;
strncpy(config.aParamFilePath, ParamFilePath, sizeof(config.aParamFilePath));
RK_MPI_AO_SetVqeAttr(m_channel, &config);
RK_MPI_AO_EnableVqe(m_channel);
}
return true;
}
void Output::close() {
if (m_channel >= 0) {
RK_MPI_AO_DisableVqe(m_channel);
RK_MPI_AO_DisableChn(m_channel);
m_channel = -1;
}
}
void Output::write(const uint8_t *data, uint32_t byteSize) {
if (m_channel < 0) return;
auto buffer = RK_MPI_MB_CreateAudioBuffer(byteSize, RK_FALSE);
if (buffer != nullptr) {
memcpy(RK_MPI_MB_GetPtr(buffer), data, byteSize);
RK_MPI_MB_SetSize(buffer, byteSize);
RK_MPI_SYS_SendMediaBuffer(RK_ID_AO, m_channel, buffer);
RK_MPI_MB_ReleaseBuffer(buffer);
} else {
LOG(error) << "RK_MPI_MB_CreateAudioBuffer() failed.";
}
}
} // namespace RkAudio
PcmStreamBuffer::PcmStreamBuffer(uint32_t sampleRate, uint32_t channels, RkAudio::Format::SampleType sampleType, uint32_t popDuration,
uint32_t capacity)
: m_sampleRate(sampleRate), m_channels(channels), m_buffer(capacity), m_capacity(capacity) {
if (sampleType == RkAudio::Format::SignedInt16) {
m_pointByteSize = 2;
}
uint32_t frameSize = sampleRate * channels * m_pointByteSize * popDuration / 1000;
m_popBuffer = std::vector<uint8_t>(frameSize);
m_popFrame.data = m_popBuffer.data();
}
bool PcmStreamBuffer::push(const RkAudio::Frame &frame) {
std::lock_guard<std::mutex> locker(m_mutex);
uint32_t byteSize = availableByteSize();
uint32_t freeSize = m_capacity - byteSize;
if (freeSize < frame.byteSize) {
LOG_FORMAT(warning, "buffer is full, capacity: %d, free size: %d, need size: %d", m_capacity, freeSize, frame.byteSize);
return false;
}
if ((m_tail + frame.byteSize) > m_capacity) {
uint32_t size1 = m_capacity - m_tail;
memcpy(m_buffer.data() + m_tail, frame.data, size1);
uint32_t size2 = frame.byteSize - size1;
memcpy(m_buffer.data(), frame.data + size1, size2);
} else {
memcpy(m_buffer.data() + m_tail, frame.data, frame.byteSize);
}
m_tail = (m_tail + frame.byteSize) % m_capacity;
m_full = (m_tail == m_head);
return true;
}
std::chrono::milliseconds PcmStreamBuffer::availableDuration() const {
auto byteSize = availableByteSize();
return std::chrono::milliseconds(1000 * byteSize / (m_sampleRate * m_channels * m_pointByteSize));
}
const RkAudio::Frame *PcmStreamBuffer::pop() {
// return nullptr;
std::lock_guard<std::mutex> locker(m_mutex);
auto byteSize = availableByteSize();
if (byteSize < m_popBuffer.size()) return nullptr;
if ((m_head + m_popBuffer.size()) > m_capacity) {
uint32_t size1 = m_capacity - m_head;
memcpy(m_popBuffer.data(), m_buffer.data() + m_head, size1);
uint32_t size2 = m_popBuffer.size() - size1;
memcpy(m_popBuffer.data() + size1, m_buffer.data(), size2);
} else {
memcpy(m_popBuffer.data(), m_buffer.data() + m_head, m_popBuffer.size());
}
m_head = (m_head + m_popBuffer.size()) % m_capacity;
m_popFrame.byteSize = m_popBuffer.size();
m_popFrame.frameSize = m_popFrame.byteSize / m_channels / m_pointByteSize;
m_full = false;
return &m_popFrame;
}
uint32_t PcmStreamBuffer::availableByteSize() const {
if (m_full) {
return m_capacity;
} else if (m_tail >= m_head) {
return m_tail - m_head;
} else {
return m_capacity + m_tail - m_head;
}
}