Decoding pcm_s16le with FFMPEG?

Question

i have a problem decoding a wav file using ffmpeg. I'm new to it and i'm not quite used to it.

In my application i have to input the audio file and get an array of samples to work on. I used ffmpeg to create a function that gets in input the path of the file, the position in time where to start to output the samples and the lenght of the chunk to decode in seconds.

I have no reputation, so I had to make a gdrive directory where you can see the problem and the files on which I worked.

Here it is: https://goo.gl/8KnjAj

When I try to decode the file harp.wav everything runs fine, and I can plot the samples as in the image plot-harp.png

The file is a WAV file encoded as: pcm_u8, 11025 Hz, 1 channels, u8, 88 kb/s

The problems comes when i try to decode the file demo-unprocessed.wav. It outputs a series of samples that has no sense. It outputs a serie of samples plotted as the image graph1-demo.jpg shows.

The file is a WAV file encoded as: pcm_s16le, 44100 Hz, 1 channels, s16, 705 kb/s

IDK where the problem in my code is, I already checked the code before and after the decoding with FFMPEG, and it works absolutely fine.

Here is the code for the dataReader.cpp :

/* Start by including the necessary */
#include "dataReader.h"
#include <cstdlib>
#include <iostream>
#include <fstream>

#ifdef __cplusplus
extern "C" {
#endif
    #include <libavcodec/avcodec.h> 
    #include <libavformat/avformat.h>
    #include <libavutil/avutil.h>
#ifdef __cplusplus 
}
#endif

using namespace std;

/* initialization function for audioChunk */
audioChunk::audioChunk(){
    data=NULL;
    size=0;
    bitrate=0;
}

/* function to get back chunk lenght in seconds */
int audioChunk::getTimeLenght(){
    return size/bitrate;
}

/* initialization function for audioChunk_dNorm */
audioChunk_dNorm::audioChunk_dNorm(){
    data=NULL;
    size=0;
    bitrate=0;
}

/* function to get back chunk lenght in seconds */
int audioChunk_dNorm::getTimeLenght(){
    return size/bitrate;
}

/* function to normalize audioChunk into audioChunk_dNorm */
void audioChunk_dNorm::fillAudioChunk(audioChunk* cnk){

    size=cnk->size;
    bitrate=cnk->bitrate;

    double min=cnk->data[0];
    double max=cnk->data[0];

    for(int i=0;i<cnk->size;i++){
        if(*(cnk->data+i)>max) max=*(cnk->data+i);
        else if(*(cnk->data+i)<min) min=*(cnk->data+i);
    }

    data=new double[size];

    for(int i=0;i<size;i++){
        //data[i]=cnk->data[i]+256*data[i+1];
        if(data[i]!=255) data[i]=2*((cnk->data[i])-(max-min)/2)/(max-min);
        else data[i]=0;
    }
    cout<<"bitrate "<<bitrate<<endl;
}


audioChunk readData(const char* path_name, const double start_time, const double lenght){

    /* inizialize audioChunk */
    audioChunk output;

    /* Check input times */
    if((start_time<0)||(lenght<0)) {
        cout<<"Input times should be positive";
        return output;
    }

    /* Start FFmpeg */
    av_register_all();

    /* Initialize the frame to read the data and verify memory allocation */
    AVFrame* frame = av_frame_alloc();
    if (!frame)
    {
        cout << "Error allocating the frame" << endl;
        return output;
    }

    /* Initialization of the Context, to open the file */
    AVFormatContext* formatContext = NULL;
    /* Opening the file, and check if it has opened */
    if (avformat_open_input(&formatContext, path_name, NULL, NULL) != 0)
    {
        av_frame_free(&frame);
        cout << "Error opening the file" << endl;
        return output;
    }

    /* Find the stream info, if not found, exit */
    if (avformat_find_stream_info(formatContext, NULL) < 0)
    {
        av_frame_free(&frame);
        avformat_close_input(&formatContext);
        cout << "Error finding the stream info" << endl;
        return output;
    }

    /* Check inputs to verify time input */
    if(start_time>(formatContext->duration/1000000)){
        cout<< "Error, start_time is over file duration"<<endl;
        av_frame_free(&frame);
        avformat_close_input(&formatContext);
        return output;
    }

    /* Chunk = number of samples to output */
    long long int chunk = ((formatContext->bit_rate)*lenght/8);
    /* Start = address of sample where start to read */
    long long int start = ((formatContext->bit_rate)*start_time/8);
    /* Tot_sampl = number of the samples in the file */
    long long int tot_sampl = (formatContext->bit_rate)*(formatContext->duration)/8000000;

    /* Set the lenght of chunk to avoid segfault and to read all the file */
    if (start+chunk>tot_sampl) {chunk = tot_sampl-start;}
    if (lenght==0) {start = 0; chunk = tot_sampl;}

    /* initialize the array to output */
    output.data = new unsigned char[chunk];
    output.bitrate = formatContext->bit_rate;
    output.size=chunk;

    av_dump_format(formatContext,0,NULL,0);
    cout<<chunk<<" n of sample to read"<<endl;
    cout<<start<<" start"<<endl;
    cout<<output.bitrate<<" bitrate"<<endl;
    cout<<tot_sampl<<" total sample"<<endl;


    /* Find the audio Stream, if no audio stream are found, clean and exit */
    AVCodec* cdc = NULL;
    int streamIndex = av_find_best_stream(formatContext, AVMEDIA_TYPE_AUDIO, -1, -1, &cdc, 0);
    if (streamIndex < 0)
    {
        av_frame_free(&frame);
        avformat_close_input(&formatContext);
        cout << "Could not find any audio stream in the file" << endl;
        return output;
    }

    /* Open the audio stream to read data  in audioStream */
    AVStream* audioStream = formatContext->streams[streamIndex];

    /* Initialize the codec context */
    AVCodecContext* codecContext = audioStream->codec;
    codecContext->codec = cdc;
    /* Open the codec, and verify if it has opened */
    if (avcodec_open2(codecContext, codecContext->codec, NULL) != 0)
    {
        av_frame_free(&frame);
        avformat_close_input(&formatContext);
        cout << "Couldn't open the context with the decoder" << endl;
        return output;
    }

    /* Initialize buffer to store compressed packets */
    AVPacket readingPacket;
    av_init_packet(&readingPacket);


    int j=0;
    int count = 0; 

    while(av_read_frame(formatContext, &readingPacket)==0){
        if((count+readingPacket.size)>start){
            if(readingPacket.stream_index == audioStream->index){

                AVPacket decodingPacket = readingPacket;

                // Audio packets can have multiple audio frames in a single packet
                while (decodingPacket.size > 0){
                    // Try to decode the packet into a frame
                    // Some frames rely on multiple packets, so we have to make sure the frame is finished before
                    // we can use it
                    int gotFrame = 0;
                    int result = avcodec_decode_audio4(codecContext, frame, &gotFrame, &decodingPacket);

                    count += result;

                    if (result >= 0 && gotFrame)
                    {
                        decodingPacket.size -= result;
                        decodingPacket.data += result;
                        int a;

                        for(int i=0;i<result-1;i++){

                            *(output.data+j)=frame->data[0][i];

                            j++;
                            if(j>=chunk) break;
                        }

                        // We now have a fully decoded audio frame
                    }
                    else
                    {
                        decodingPacket.size = 0;
                        decodingPacket.data = NULL;
                    }
                    if(j>=chunk) break;
                }
            }              
        }else count+=readingPacket.size;

        // To prevent memory leak, must free packet.
        av_free_packet(&readingPacket);
        if(j>=chunk) break;
    }

    // Some codecs will cause frames to be buffered up in the decoding process. If the CODEC_CAP_DELAY flag
    // is set, there can be buffered up frames that need to be flushed, so we'll do that
    if (codecContext->codec->capabilities & CODEC_CAP_DELAY)
    {
        av_init_packet(&readingPacket);
        // Decode all the remaining frames in the buffer, until the end is reached
        int gotFrame = 0;
        int a;
        int result=avcodec_decode_audio4(codecContext, frame, &gotFrame, &readingPacket);
        while (result >= 0 && gotFrame)
        {
            // We now have a fully decoded audio frame
            for(int i=0;i<result-1;i++){

                *(output.data+j)=frame->data[0][i];

                j++;
                if(j>=chunk) break;
            }
            if(j>=chunk) break;
        }
    }

    // Clean up!
    av_free(frame);
    avcodec_close(codecContext);
    avformat_close_input(&formatContext);

    cout<<"Ended Reading, "<<j<<" samples read"<<endl;
    output.size=j;
    return output;
}

Here is the dataReader.h

/* 
 * File:   dataReader.h
 * Author: davide
 *
 * Created on 27 luglio 2015, 11.11
 */

#ifndef DATAREADER_H
#define DATAREADER_H

/* function that reads a file and outputs an array of samples
 * @ path_name = the path of the file to read
 * @ start_time = the position where to start the data reading, 0 = start
 *                the time is in seconds, it can hold to 10e-6 seconds
 * @ lenght = the lenght of the frame to extract the data, 
 *            0 = read all the file (do not use with big files)
 *            if lenght > of file duration, it reads through the end of file.
 *            the time is in seconds, it can hold to 10e-6 seconds  
 */

#include <stdint.h>

class audioChunk{
public:
    uint8_t *data;
    unsigned int size;
    int bitrate;
    int getTimeLenght();
    audioChunk();
};

class audioChunk_dNorm{
public:
    double* data;
    unsigned int size;
    int bitrate;
    int getTimeLenght();
    void fillAudioChunk(audioChunk* cnk);
    audioChunk_dNorm();
};

audioChunk readData(const char* path_name, const double start_time, const double lenght);

#endif  /* DATAREADER_H */

And finally there is the main.cpp of the application.

/* 
 * File:   main.cpp
 * Author: davide
 *
 * Created on 28 luglio 2015, 17.04
 */

#include <cstdlib>
#include "dataReader.h"
#include "transforms.h"
#include "tognuplot.h"
#include <fstream>
#include <iostream>

using namespace std;

/*
 * 
 */
int main(int argc, char** argv) {

    audioChunk *chunk1=new audioChunk;

    audioChunk_dNorm *normChunk1=new audioChunk_dNorm;

    *chunk1=readData("./audio/demo-unprocessed.wav",0,1);

    normChunk1->fillAudioChunk(chunk1);

    ofstream file1;
    file1.open("./file/2wave.txt", std::ofstream::trunc);
    if(file1.is_open()) {
        for(int i=0;i<chunk1->size;i++) {
            int a=chunk1->data[i];
            file1<<i<<" "<<a<<endl;
        }
    }
    else cout<<"Error opening file";

    file1.close();

    return 0;
}

I can't understand why the outputs goes like this. Is it possible that the decoder can't convert the samples (pcm_16le, 16bits) into FFMPEG AVFrame.data, that stores the samples ad uint8_t? And if it is it is there some way to make FFMPEG work for audio files that stores samples at more than 8 bits?

The file graph1-demo_good.jpg is how the samples should be, extracted with a working LIBSNDFILE application that I made.

EDIT: Seems like the program can't convert the decoded data, couples of little endian bytes stored in a couple of uint8_t unsigned char, into the destination format (that i set as unsigned char[]), because it stores the bits as little-endian 16 bytes. So the data into audioChunk.data is right, but I have to read it not as an unsigned char, but as a couple of little-endian bytes.

Answer 1

I looked at the memory pointed to by chunk1->data using gdb. (x /256xh 0x18dddf0, to dump the first 256 half-words in hex). It looks like signed 16bit values, since it starts off with a lot of 0, 0xFFFF, and 0x0001.

So your code needs to ask ffmpeg to convert to a specific format. IDK how best to do that, sorry.