| | |
| | | #include "audio.h" |
| | | #include "precomp.h" |
| | | |
| | | extern "C" { |
| | | #include <libavutil/opt.h> |
| | | #include <libavcodec/avcodec.h> |
| | | #include <libavformat/avformat.h> |
| | | #include <libavutil/channel_layout.h> |
| | | #include <libavutil/samplefmt.h> |
| | | #include <libswresample/swresample.h> |
| | | } |
| | | |
| | | using namespace std; |
| | | |
| | | namespace funasr { |
| | |
| | | memset(speech_data, 0, sizeof(float) * speech_len); |
| | | copy(samples.begin(), samples.end(), speech_data); |
| | | } |
| | | |
| | | bool Audio::FfmpegLoad(const char *filename){ |
| | | // from file |
| | | AVFormatContext* formatContext = avformat_alloc_context(); |
| | | if (avformat_open_input(&formatContext, filename, NULL, NULL) != 0) { |
| | | printf("Error: Could not open input file."); |
| | | avformat_close_input(&formatContext); |
| | | avformat_free_context(formatContext); |
| | | return false; |
| | | } |
| | | |
| | | if (avformat_find_stream_info(formatContext, NULL) < 0) { |
| | | printf("Error: Could not find stream information."); |
| | | avformat_close_input(&formatContext); |
| | | avformat_free_context(formatContext); |
| | | return false; |
| | | } |
| | | const AVCodec* codec = NULL; |
| | | AVCodecParameters* codecParameters = NULL; |
| | | int audioStreamIndex = av_find_best_stream(formatContext, AVMEDIA_TYPE_AUDIO, -1, -1, &codec, 0); |
| | | if (audioStreamIndex >= 0) { |
| | | codecParameters = formatContext->streams[audioStreamIndex]->codecpar; |
| | | } |
| | | AVCodecContext* codecContext = avcodec_alloc_context3(codec); |
| | | if (!codecContext) { |
| | | fprintf(stderr, "Failed to allocate codec context\n"); |
| | | avformat_close_input(&formatContext); |
| | | avformat_free_context(formatContext); |
| | | return false; |
| | | } |
| | | if (avcodec_parameters_to_context(codecContext, codecParameters) != 0) { |
| | | printf("Error: Could not copy codec parameters to codec context."); |
| | | avformat_close_input(&formatContext); |
| | | avformat_free_context(formatContext); |
| | | avcodec_free_context(&codecContext); |
| | | return false; |
| | | } |
| | | if (avcodec_open2(codecContext, codec, NULL) < 0) { |
| | | printf("Error: Could not open audio decoder."); |
| | | avformat_close_input(&formatContext); |
| | | avformat_free_context(formatContext); |
| | | avcodec_free_context(&codecContext); |
| | | return false; |
| | | } |
| | | SwrContext *swr_ctx = swr_alloc_set_opts( |
| | | nullptr, // allocate a new context |
| | | AV_CH_LAYOUT_MONO, // output channel layout (stereo) |
| | | AV_SAMPLE_FMT_S16, // output sample format (signed 16-bit) |
| | | 16000, // output sample rate (same as input) |
| | | av_get_default_channel_layout(codecContext->channels), // input channel layout |
| | | codecContext->sample_fmt, // input sample format |
| | | codecContext->sample_rate, // input sample rate |
| | | 0, // logging level |
| | | nullptr // parent context |
| | | ); |
| | | if (swr_ctx == nullptr) { |
| | | std::cerr << "Could not initialize resampler" << std::endl; |
| | | avformat_close_input(&formatContext); |
| | | avformat_free_context(formatContext); |
| | | avcodec_free_context(&codecContext); |
| | | return false; |
| | | } |
| | | if (swr_init(swr_ctx) != 0) { |
| | | std::cerr << "Could not initialize resampler" << std::endl; |
| | | avformat_close_input(&formatContext); |
| | | avformat_free_context(formatContext); |
| | | avcodec_free_context(&codecContext); |
| | | swr_free(&swr_ctx); |
| | | return false; |
| | | } |
| | | |
| | | // to pcm |
| | | AVPacket* packet = av_packet_alloc(); |
| | | AVFrame* frame = av_frame_alloc(); |
| | | std::vector<uint8_t> resampled_buffers; |
| | | while (av_read_frame(formatContext, packet) >= 0) { |
| | | if (packet->stream_index == audioStreamIndex) { |
| | | if (avcodec_send_packet(codecContext, packet) >= 0) { |
| | | while (avcodec_receive_frame(codecContext, frame) >= 0) { |
| | | // Resample audio if necessary |
| | | std::vector<uint8_t> resampled_buffer; |
| | | int in_samples = frame->nb_samples; |
| | | uint8_t **in_data = frame->extended_data; |
| | | int out_samples = av_rescale_rnd(in_samples, |
| | | 16000, |
| | | codecContext->sample_rate, |
| | | AV_ROUND_DOWN); |
| | | |
| | | int resampled_size = out_samples * av_get_bytes_per_sample(AV_SAMPLE_FMT_S16); |
| | | if (resampled_buffer.size() < resampled_size) { |
| | | resampled_buffer.resize(resampled_size); |
| | | } |
| | | uint8_t *resampled_data = resampled_buffer.data(); |
| | | int ret = swr_convert( |
| | | swr_ctx, |
| | | &resampled_data, // output buffer |
| | | resampled_size, // output buffer size |
| | | (const uint8_t **)(frame->data), //(const uint8_t **)(frame->extended_data) |
| | | in_samples // input buffer size |
| | | ); |
| | | if (ret < 0) { |
| | | std::cerr << "Error resampling audio" << std::endl; |
| | | break; |
| | | } |
| | | std::copy(resampled_buffer.begin(), resampled_buffer.end(), std::back_inserter(resampled_buffers)); |
| | | } |
| | | } |
| | | } |
| | | av_packet_unref(packet); |
| | | } |
| | | |
| | | avformat_close_input(&formatContext); |
| | | avformat_free_context(formatContext); |
| | | avcodec_free_context(&codecContext); |
| | | swr_free(&swr_ctx); |
| | | av_packet_free(&packet); |
| | | av_frame_free(&frame); |
| | | |
| | | if (speech_data != NULL) { |
| | | free(speech_data); |
| | | } |
| | | if (speech_buff != NULL) { |
| | | free(speech_buff); |
| | | } |
| | | offset = 0; |
| | | |
| | | speech_len = (resampled_buffers.size()) / 2; |
| | | speech_buff = (int16_t*)malloc(sizeof(int16_t) * speech_len); |
| | | if (speech_buff) |
| | | { |
| | | memset(speech_buff, 0, sizeof(int16_t) * speech_len); |
| | | memcpy((void*)speech_buff, (const void*)resampled_buffers.data(), speech_len * sizeof(int16_t)); |
| | | |
| | | speech_data = (float*)malloc(sizeof(float) * speech_len); |
| | | memset(speech_data, 0, sizeof(float) * speech_len); |
| | | |
| | | float scale = 1; |
| | | if (data_type == 1) { |
| | | scale = 32768; |
| | | } |
| | | for (int32_t i = 0; i != speech_len; ++i) { |
| | | speech_data[i] = (float)speech_buff[i] / scale; |
| | | } |
| | | |
| | | AudioFrame* frame = new AudioFrame(speech_len); |
| | | frame_queue.push(frame); |
| | | |
| | | return true; |
| | | } |
| | | else |
| | | return false; |
| | | |
| | | } |
| | | |
| | | bool Audio::FfmpegLoad(const char* buf, int n_file_len){ |
| | | // from buf |
| | | char* buf_copy = (char *)malloc(n_file_len); |
| | | memcpy(buf_copy, buf, n_file_len); |
| | | |
| | | AVIOContext* avio_ctx = avio_alloc_context( |
| | | (unsigned char*)buf_copy, // buffer |
| | | n_file_len, // buffer size |
| | | 0, // write flag (0 for read-only) |
| | | nullptr, // opaque pointer (not used here) |
| | | nullptr, // read callback (not used here) |
| | | nullptr, // write callback (not used here) |
| | | nullptr // seek callback (not used here) |
| | | ); |
| | | AVFormatContext* formatContext = avformat_alloc_context(); |
| | | formatContext->pb = avio_ctx; |
| | | if (avformat_open_input(&formatContext, "", NULL, NULL) != 0) { |
| | | printf("Error: Could not open input file."); |
| | | avio_context_free(&avio_ctx); |
| | | avformat_close_input(&formatContext); |
| | | avformat_free_context(formatContext); |
| | | return false; |
| | | } |
| | | |
| | | if (avformat_find_stream_info(formatContext, NULL) < 0) { |
| | | printf("Error: Could not find stream information."); |
| | | avio_context_free(&avio_ctx); |
| | | avformat_close_input(&formatContext); |
| | | avformat_free_context(formatContext); |
| | | return false; |
| | | } |
| | | const AVCodec* codec = NULL; |
| | | AVCodecParameters* codecParameters = NULL; |
| | | int audioStreamIndex = av_find_best_stream(formatContext, AVMEDIA_TYPE_AUDIO, -1, -1, &codec, 0); |
| | | if (audioStreamIndex >= 0) { |
| | | codecParameters = formatContext->streams[audioStreamIndex]->codecpar; |
| | | } |
| | | AVCodecContext* codecContext = avcodec_alloc_context3(codec); |
| | | if (!codecContext) { |
| | | fprintf(stderr, "Failed to allocate codec context\n"); |
| | | avio_context_free(&avio_ctx); |
| | | avformat_close_input(&formatContext); |
| | | avformat_free_context(formatContext); |
| | | return false; |
| | | } |
| | | if (avcodec_parameters_to_context(codecContext, codecParameters) != 0) { |
| | | printf("Error: Could not copy codec parameters to codec context."); |
| | | avio_context_free(&avio_ctx); |
| | | avformat_close_input(&formatContext); |
| | | avformat_free_context(formatContext); |
| | | avcodec_free_context(&codecContext); |
| | | return false; |
| | | } |
| | | if (avcodec_open2(codecContext, codec, NULL) < 0) { |
| | | printf("Error: Could not open audio decoder."); |
| | | avio_context_free(&avio_ctx); |
| | | avformat_close_input(&formatContext); |
| | | avformat_free_context(formatContext); |
| | | avcodec_free_context(&codecContext); |
| | | return false; |
| | | } |
| | | SwrContext *swr_ctx = swr_alloc_set_opts( |
| | | nullptr, // allocate a new context |
| | | AV_CH_LAYOUT_MONO, // output channel layout (stereo) |
| | | AV_SAMPLE_FMT_S16, // output sample format (signed 16-bit) |
| | | 16000, // output sample rate (same as input) |
| | | av_get_default_channel_layout(codecContext->channels), // input channel layout |
| | | codecContext->sample_fmt, // input sample format |
| | | codecContext->sample_rate, // input sample rate |
| | | 0, // logging level |
| | | nullptr // parent context |
| | | ); |
| | | if (swr_ctx == nullptr) { |
| | | std::cerr << "Could not initialize resampler" << std::endl; |
| | | avio_context_free(&avio_ctx); |
| | | avformat_close_input(&formatContext); |
| | | avformat_free_context(formatContext); |
| | | avcodec_free_context(&codecContext); |
| | | return false; |
| | | } |
| | | if (swr_init(swr_ctx) != 0) { |
| | | std::cerr << "Could not initialize resampler" << std::endl; |
| | | avio_context_free(&avio_ctx); |
| | | avformat_close_input(&formatContext); |
| | | avformat_free_context(formatContext); |
| | | avcodec_free_context(&codecContext); |
| | | swr_free(&swr_ctx); |
| | | return false; |
| | | } |
| | | |
| | | // to pcm |
| | | AVPacket* packet = av_packet_alloc(); |
| | | AVFrame* frame = av_frame_alloc(); |
| | | std::vector<uint8_t> resampled_buffers; |
| | | while (av_read_frame(formatContext, packet) >= 0) { |
| | | if (packet->stream_index == audioStreamIndex) { |
| | | if (avcodec_send_packet(codecContext, packet) >= 0) { |
| | | while (avcodec_receive_frame(codecContext, frame) >= 0) { |
| | | // Resample audio if necessary |
| | | std::vector<uint8_t> resampled_buffer; |
| | | int in_samples = frame->nb_samples; |
| | | uint8_t **in_data = frame->extended_data; |
| | | int out_samples = av_rescale_rnd(in_samples, |
| | | 16000, |
| | | codecContext->sample_rate, |
| | | AV_ROUND_DOWN); |
| | | |
| | | int resampled_size = out_samples * av_get_bytes_per_sample(AV_SAMPLE_FMT_S16); |
| | | if (resampled_buffer.size() < resampled_size) { |
| | | resampled_buffer.resize(resampled_size); |
| | | } |
| | | uint8_t *resampled_data = resampled_buffer.data(); |
| | | int ret = swr_convert( |
| | | swr_ctx, |
| | | &resampled_data, // output buffer |
| | | resampled_size, // output buffer size |
| | | (const uint8_t **)(frame->data), //(const uint8_t **)(frame->extended_data) |
| | | in_samples // input buffer size |
| | | ); |
| | | if (ret < 0) { |
| | | std::cerr << "Error resampling audio" << std::endl; |
| | | break; |
| | | } |
| | | std::copy(resampled_buffer.begin(), resampled_buffer.end(), std::back_inserter(resampled_buffers)); |
| | | } |
| | | } |
| | | } |
| | | av_packet_unref(packet); |
| | | } |
| | | |
| | | avio_context_free(&avio_ctx); |
| | | avformat_close_input(&formatContext); |
| | | avformat_free_context(formatContext); |
| | | avcodec_free_context(&codecContext); |
| | | swr_free(&swr_ctx); |
| | | av_packet_free(&packet); |
| | | av_frame_free(&frame); |
| | | |
| | | if (speech_data != NULL) { |
| | | free(speech_data); |
| | | } |
| | | if (speech_buff != NULL) { |
| | | free(speech_buff); |
| | | } |
| | | offset = 0; |
| | | |
| | | speech_len = (resampled_buffers.size()) / 2; |
| | | speech_buff = (int16_t*)malloc(sizeof(int16_t) * speech_len); |
| | | if (speech_buff) |
| | | { |
| | | memset(speech_buff, 0, sizeof(int16_t) * speech_len); |
| | | memcpy((void*)speech_buff, (const void*)resampled_buffers.data(), speech_len * sizeof(int16_t)); |
| | | |
| | | speech_data = (float*)malloc(sizeof(float) * speech_len); |
| | | memset(speech_data, 0, sizeof(float) * speech_len); |
| | | |
| | | float scale = 1; |
| | | if (data_type == 1) { |
| | | scale = 32768; |
| | | } |
| | | for (int32_t i = 0; i != speech_len; ++i) { |
| | | speech_data[i] = (float)speech_buff[i] / scale; |
| | | } |
| | | |
| | | AudioFrame* frame = new AudioFrame(speech_len); |
| | | frame_queue.push(frame); |
| | | |
| | | return true; |
| | | } |
| | | else |
| | | return false; |
| | | |
| | | } |
| | | |
| | | |
| | | bool Audio::LoadWav(const char *filename, int32_t* sampling_rate) |
| | | { |
| | |
| | | return true; |
| | | } |
| | | |
| | | bool Audio::LoadOthers2Char(const char* filename) |
| | | { |
| | | if (speech_char != NULL) { |
| | | free(speech_char); |
| | | } |
| | | |
| | | FILE* fp; |
| | | fp = fopen(filename, "rb"); |
| | | if (fp == nullptr) |
| | | { |
| | | LOG(ERROR) << "Failed to read " << filename; |
| | | return false; |
| | | } |
| | | fseek(fp, 0, SEEK_END); |
| | | uint32_t n_file_len = ftell(fp); |
| | | fseek(fp, 0, SEEK_SET); |
| | | |
| | | speech_len = n_file_len; |
| | | speech_char = (char *)malloc(n_file_len); |
| | | memset(speech_char, 0, n_file_len); |
| | | fread(speech_char, 1, n_file_len, fp); |
| | | fclose(fp); |
| | | |
| | | return true; |
| | | } |
| | | |
| | | int Audio::FetchChunck(float *&dout, int len) |
| | | { |
| | | if (offset >= speech_align_len) { |