コード例 #1
0
ファイル: AAudioAudio.cpp プロジェクト: MIPS/frameworks-av
AAUDIO_API aaudio_result_t AAudioStream_read(AAudioStream* stream,
                               void *buffer,
                               int32_t numFrames,
                               int64_t timeoutNanoseconds)
{
    AudioStream *audioStream = convertAAudioStreamToAudioStream(stream);
    if (buffer == nullptr) {
        return AAUDIO_ERROR_NULL;
    }
    if (numFrames < 0) {
        return AAUDIO_ERROR_ILLEGAL_ARGUMENT;
    } else if (numFrames == 0) {
        return 0;
    }

    aaudio_result_t result = audioStream->read(buffer, numFrames, timeoutNanoseconds);

    return result;
}
コード例 #2
0
int main(int argc, char *argv[]) {

    init(argc, argv);

    AudioFile a(input1.c_str());
    AudioFile b(input2.c_str());

    if (channel >= a.getNumberOfChannels() || channel >= b.getNumberOfChannels()) {
		ls << log_error() << "Error, channel not available" << endl;
		return 1;
    }

    //cross correlate x seconds at a time.

    uint32_t samplesPrBlock = a.getSampleRate() * secondsPrBlock;
    ls << log_debug() << "samples a: " << samplesPrBlock << endl;
    ls << log_debug() << "samples b: " << b.getSampleRate() * secondsPrBlock << endl;
    AudioStream aStream = a.getStream(channel);
    AudioStream bStream = b.getStream(channel);

    bool done = false;
    bool first = true;
    bool success = true;
    size_t blockFailure = 0;

    double blockFailureVal = 0.0;
    int64_t blockFailureOffset = 0;

    double firstMaxVal = 0.0;
    int64_t firstOffset = 0;

    double minimumVal = 2;

    for (size_t block = 1; !done; ++block) {
        vector<int16_t> aSamples, bSamples;
        vector<uint64_t> aSquarePrefixSum, bSquarePrefixSum;
        vector<complex<double> > result;

        aStream.read(samplesPrBlock, aSamples);
        bStream.read(samplesPrBlock, bSamples);

        if (aSamples.size() == 0 && bSamples.size() == 0) {
            // reached the end of both of the samples
            if (verbose) {
                cout << "finished: reached end of both samples" << endl;
            }
            break;
        }

        if (padShortBlock) {
            // If this correlation involves 'short' files, shorter than one
            // block, pad the blocks with silence - this will only pad the last
            // block, so that for short files, shorter than one block, the
            // correlation is performed, rather than terminating the loop early
            // in the following break, leaving minimumVal as 2, and indicating
            // success.  Because we're looping until the end of both samples,
            // this will pad out a shorter sample with silence (which will
            // probably yield a correlation failure)
            aSamples.resize(samplesPrBlock, 0);
            bSamples.resize(samplesPrBlock, 0);
        }

        if (aSamples.size() < samplesPrBlock/2 || bSamples.size() < samplesPrBlock) {
            // not enough samples for another reliable check.
            if (verbose) {
                cout << "finished: not enough samples for another reliable check" << endl;
            }
            break;
        }

        prefixSquareSum(aSamples, aSquarePrefixSum);
        prefixSquareSum(bSamples, bSquarePrefixSum);

        // we count the average of absolute values
        // if the average is close to 0, then we decide it is silence

        size_t absSumA = 0; size_t absSumB = 0;
        bool silence = false;
        for (size_t i = 0; i < aSamples.size(); ++i) {
            absSumA += (aSamples[i]>=0)?aSamples[i]:-aSamples[i];
        }
        for (size_t i = 0; i < bSamples.size(); ++i) {
            absSumB += (bSamples[i]>=0)?bSamples[i]:-bSamples[i];
        }

        double avgA = static_cast<double>(absSumA)/aSamples.size();
        double avgB = static_cast<double>(absSumB)/bSamples.size();

        if (avgA <= 5.0 && avgB <= 5.0) {
            silence = true;
        } else if (avgA <= 5.0 || avgB <= 5.0) {
            success = false;
        }

        bool compare = !silence;

        int64_t maxIdx = -1; double maxVal = -1.0;

        if (compare) {
            proxyFFT<int16_t, double> aFFT(aSamples);
            proxyFFT<int16_t, double> bFFT(bSamples);

            pair<int64_t, double> tmp = compareBlock(aFFT, bFFT, aSquarePrefixSum, bSquarePrefixSum);

            maxIdx = tmp.first;
            maxVal = tmp.second;

            tmp = compareBlock(bFFT, aFFT, bSquarePrefixSum, aSquarePrefixSum);

            if (tmp.second > maxVal + 1e-6) {
                maxIdx = -tmp.first;
                maxVal = tmp.second;
            }

            if (first) {
                first = false;
                firstMaxVal = maxVal;
                firstOffset = maxIdx;
                minimumVal = maxVal;
                if (firstMaxVal < threshold) {
                    if (verbose) {
                        cout << "failed: threshold crossed in first block (" << firstMaxVal << ")" << endl;
                    }
                    success = false;
                    blockFailure = block;
                    blockFailureVal = maxVal;
                    blockFailureOffset = maxIdx;
                }
            } else {
                int64_t offsetDistance = abs(maxIdx - firstOffset);
                bool offsetDistanceExceeded = offsetDistance > 500;
                bool thresholdCrossed = maxVal < threshold;
                if (offsetDistanceExceeded || thresholdCrossed) {
                    if (verbose && offsetDistanceExceeded) {
                        cout << "failed: offset distance exceeded (" << offsetDistance << ")" << endl;
                    }
                    if (verbose && thresholdCrossed) {
                        cout << "failed: threshold crossed (" << maxVal << ")" << endl;
                    }
                    // check to see that the offset between blocks is not too large.
                    success = false;
                    blockFailure = block;
                    blockFailureVal = maxVal;
                    blockFailureOffset = maxIdx;
                }
                if (maxVal < minimumVal) {
                    minimumVal = maxVal;
                }
            }
        }

        if (aSamples.size() < samplesPrBlock || bSamples.size() < samplesPrBlock) {
            // we don't check the last block. Is this ok?
            if (verbose) {
                cout << "finished: at last block" << endl;
            }
            done = true;
        }
        if (verbose) {
            if (compare) {
                cout << "block " << block << ": " << maxVal << " " << maxIdx << endl;
            } else if (silence) {
                cout << "block " << block << ": " << "silence" << endl;
            }
        }
    }

    if (success) {
      cout << "Success" << endl;
      cout << "Offset: " << firstOffset << endl;
      cout << "Similarity: " << minimumVal << endl;
      return 0;
    } else {
      cout << "Failure" << endl;
      cout << "Block: " << blockFailure << endl;
      cout << "Value in block: " << blockFailureVal << endl;
      cout << "Offset in block: " << blockFailureOffset << " (normal: " << firstOffset << ")" << endl;
      return 1;
		/*
		  cout << "block " << blockFailure << ":" << endl;
		  cout << "Time: " << getTimestampFromSeconds(blockFailure*5-5) << " - "
		  << getTimestampFromSeconds(blockFailure*5) << " did not match properly" << endl;
		*/
    }

}
コード例 #3
0
/**
 * The playback thread code
 * \internal
 */
void *narrator_thread(void *narrator)
{
    //Narrator *n = static_cast<Narrator *>(narrator);
    Narrator *n = (Narrator*)narrator;

    int queueitems;

    // Set initial values to 0 so that they get updated when thread gets play signal
    float gain = 0;
    float tempo = 0;
    float pitch = 0;

    PortAudio portaudio;
    Filter filter;

    Narrator::threadState state = n->getState();
    LOG4CXX_INFO(narratorLog, "Starting playback thread");

    do {
        queueitems = n->numPlaylistItems();

        if(queueitems == 0) {
            // Wait a little before calling callback
            long waitms = portaudio.getRemainingms();
            if(waitms != 0) {
                LOG4CXX_DEBUG(narratorLog, "Waiting " << waitms << " ms for playback to finish");
                while(waitms > 0 && queueitems == 0) {
                    usleep(100000);
                    queueitems = n->numPlaylistItems();
                    waitms -= 100;
                }
            }

            // Break if we during the pause got some more queued items to play
            if(queueitems == 0) {
                if(state != Narrator::DEAD)
                    n->audioFinishedPlaying();
                n->setState(Narrator::WAIT);
                LOG4CXX_INFO(narratorLog, "Narrator in WAIT state");
                portaudio.stop();

                while(queueitems == 0) {
                    state = n->getState();
                    if(state == Narrator::EXIT) break;

                    usleep(10000);
                    queueitems = n->numPlaylistItems();
                }
            }
            LOG4CXX_INFO(narratorLog, "Narrator starting playback");
        }

        if(state == Narrator::EXIT) break;

        n->setState(Narrator::PLAY);
        n->bResetFlag = false;

        Narrator::PlaylistItem pi;

        pthread_mutex_lock(n->narratorMutex);
        if(n->mPlaylist.size() > 0) {
            pi = n->mPlaylist.front();
            n->mPlaylist.pop();
        } else {
            LOG4CXX_ERROR(narratorLog, "Narrator started playback thread without playlistitems");
            pthread_mutex_unlock(n->narratorMutex);
            continue;
        }
        string lang = n->mLanguage;
        pthread_mutex_unlock(n->narratorMutex);

        // If trying to play a file, open it
        if(pi.mClass == "file") {
            LOG4CXX_DEBUG(narratorLog, "Playing file: " << pi.mIdentifier);

            AudioStream *audioStream;

            std::string fileExtension = getFileExtension(pi.mIdentifier);
            if (fileExtension == "ogg")
            {
                audioStream = new OggStream;
            }
            else if (fileExtension == "mp3")
            {
                audioStream = new Mp3Stream;
            }
            else
            {
                LOG4CXX_ERROR(narratorLog, "extension '" << fileExtension << "' not supported");
                continue;
            }

            if(!audioStream->open(pi.mIdentifier)) {
                LOG4CXX_ERROR(narratorLog, "error opening audio stream: " << pi.mIdentifier);
                audioStream->close();
                continue;
            }

            if (portaudio.getRate() != audioStream->getRate())
            {
                long waitms = portaudio.getRemainingms();
                if (waitms != 0)
                {
                    LOG4CXX_DEBUG(narratorLog, "Waiting for current playback to finish");
                    while (waitms > 0)
                    {
                        usleep(100000);
                        waitms -= 100;
                    }
                }
            }

            if(!portaudio.open(audioStream->getRate(), audioStream->getChannels())) {
                LOG4CXX_ERROR(narratorLog, "error initializing portaudio, (rate: " << audioStream->getRate() << " channels: " << audioStream->getChannels() << ")");
                continue;
            }

            if(!filter.open(audioStream->getRate(), audioStream->getChannels())) {
                LOG4CXX_ERROR(narratorLog, "error initializing filter");
                continue;
            }

            LOG4CXX_DEBUG(narratorLog, "Audio stream has " << audioStream->getChannels() << " channel(s) and rate " << audioStream->getRate() << " Hz");

            int inSamples = 0;
            soundtouch::SAMPLETYPE* buffer = new soundtouch::SAMPLETYPE[audioStream->getChannels()*BUFFERSIZE];
            //buffer = (short*)malloc(sizeof(short) * 2 * BUFFERSIZE);
            // long totalSamplesRead = 0;
            do {
                // change gain, tempo and pitch
                adjustGainTempoPitch(n, filter, gain, tempo, pitch);

                // read some stuff from the audio stream
                inSamples = audioStream->read(buffer, BUFFERSIZE/**audioStream->getChannels()*/);
                LOG4CXX_TRACE(narratorLog, "got " << inSamples << " samples");

                //printf("Read %d samples from audio stream\n", inSamples);

                if(inSamples != 0) {
                    filter.write(buffer, inSamples); // One sample contains data for all channels here
                    writeSamplesToPortaudio( n, portaudio, filter, buffer );
                } else {
                    LOG4CXX_INFO(narratorLog, "Flushing soundtouch buffer");
                    filter.flush();
                }

                state = n->getState();

            } while (inSamples != 0 && state == Narrator::PLAY && !n->bResetFlag);

            if(buffer != NULL) delete [] (buffer);
            audioStream->close();
            delete audioStream;
        }

        // Else try opening from database
        else {
            vector <MessageAudio> vAudioQueue;

            // Get a list of MessageAudio objects to play

            Message *m = pi.mMessage;
            if(m==NULL){
                LOG4CXX_ERROR(narratorLog, "Message was null");
            }

            m->setLanguage(lang);
            m->load(pi.mIdentifier, pi.mClass);

            if(!m->compile() || !m->hasAudio()) {
                LOG4CXX_ERROR(narratorLog, "Narrator translation not found: could not find audio for '" << pi.mIdentifier << "'");
            } else {
                vAudioQueue = m->getAudioQueue();
            }

            // Play what we got
            if(vAudioQueue.size() > 0) {
                vector <MessageAudio>::iterator audio;
                audio = vAudioQueue.begin();
                do {
                    LOG4CXX_INFO(narratorLog, "Saying: " << audio->getText());

                    AudioStream *audioStream;

                    std::string encoding = ((MessageAudio&)*audio).getEncoding();
                    if (encoding == "ogg")
                    {
                        audioStream = new OggStream;
                    }
                    else if (encoding == "mp3")
                    {
                        audioStream = new Mp3Stream;
                    }
                    else
                    {
                        LOG4CXX_ERROR(narratorLog, "encoding '" << encoding << "' not supported");
                        audio++;
                        continue;
                    }

                    if(!audioStream->open(*audio)) {
                        LOG4CXX_ERROR(narratorLog, "error opening audio stream");
                        audioStream->close();
                        break;
                    }

                    if (portaudio.getRate() != audioStream->getRate())
                    {
                        long waitms = portaudio.getRemainingms();
                        if (waitms != 0)
                        {
                            LOG4CXX_DEBUG(narratorLog, "Waiting for current playback to finish");
                            while (waitms > 0)
                            {
                                usleep(100000);
                                waitms -= 100;
                            }
                        }
                    }

                    if(!portaudio.open(audioStream->getRate(), audioStream->getChannels())) {
                        LOG4CXX_ERROR(narratorLog, "error initializing portaudio");
                        break;
                    }

                    if(!filter.open(audioStream->getRate(), audioStream->getChannels())) {
                        LOG4CXX_ERROR(narratorLog, "error initializing filter");
                        break;
                    }


                    int inSamples = 0;
                    soundtouch::SAMPLETYPE* buffer = new soundtouch::SAMPLETYPE[audioStream->getChannels()*BUFFERSIZE];

                    do {
                        // change gain, tempo and pitch
                        adjustGainTempoPitch(n, filter, gain, tempo, pitch);

                        // read some stuff from the audio stream
                        inSamples = audioStream->read(buffer, BUFFERSIZE*audioStream->getChannels());

                        if(inSamples != 0) {
                            filter.write(buffer, inSamples);
                            writeSamplesToPortaudio( n, portaudio, filter, buffer );
                        } else {
                            LOG4CXX_INFO(narratorLog, "Flushing soundtouch buffer");
                            filter.flush();
                        }

                        state = n->getState();

                    } while (inSamples != 0 && state == Narrator::PLAY && !n->bResetFlag);

                    if(buffer != NULL) delete [] (buffer);
                    audioStream->close();
                    audio++;

                } while(audio != vAudioQueue.end() && state == Narrator::PLAY && !n->bResetFlag);
            }
            //Cleanup message object
            delete(pi.mMessage);
        }

        // Abort stream?
        if(n->bResetFlag) {
            n->bResetFlag = false;
            portaudio.stop();
            filter.clear();
        }

    } while(state != Narrator::EXIT);

    LOG4CXX_INFO(narratorLog, "Shutting down playbackthread");

    pthread_exit(NULL);
    return NULL;
}
コード例 #4
0
ファイル: sound_match.cpp プロジェクト: tgc/scape-xcorrsound
void match(AudioFile &needle, AudioFile &haystack, std::vector<pair<size_t, double> > &results) {
    std::vector<short> small; std::vector<short> large;
    std::vector<int64_t> smallPrefixSum; std::vector<int64_t> largePrefixSum;
    needle.getSamplesForChannel(0, small);
    prefixSquareSum(small, smallPrefixSum);

    proxyFFT<short, double> smallFFT(small);
    smallFFT.transform();
    
    size_t largeTotalSize = haystack.getNumberOfSamplesPrChannel();
    // vector<int64_t> maxSamplesBegin(largeTotalSize/small.size());
    // vector<int64_t> maxSamplesEnd(largeTotalSize/small.size());
    vector<Record> maxSamplesBegin(largeTotalSize/small.size());
    vector<Record> maxSamplesEnd(largeTotalSize/small.size());

    size_t stillToRead = largeTotalSize;

    AudioStream hayStream = haystack.getStream(0);
    size_t pieces = 13;
    for (int j = 0; ; ++j) {
	hayStream.read(pieces*small.size(), large);
	prefixSquareSum(large, largePrefixSum);
	size_t numberOfParts = large.size()/small.size();
	size_t idxAdd = j*pieces;

	// Progress information
	std::cout << '\r' << setw(8) << ((largeTotalSize-stillToRead)+0.0)/largeTotalSize*100 << " %";
	std::cout.flush();

	stillToRead -= large.size();

	for (size_t ii = 0; ii < numberOfParts*small.size(); ii += small.size()) {
	    //do stuff..
	    proxyFFT<short, double> largeFFT(large.begin()+ii, large.begin()+ii+small.size());
	    
	    vector<complex<double> > outBegin;
	    vector<complex<double> > outEnd;
	    //std::cout << "TEST1" << std::endl;
	    cross_correlation(largeFFT, smallFFT, outBegin);
	    cross_correlation(smallFFT, largeFFT, outEnd);

	    size_t maxSampleBegin = 0;
	    double maxNormFactorBegin = computeNormFactor(smallPrefixSum, largePrefixSum,
							  smallPrefixSum.begin(), smallPrefixSum.end(),
							  largePrefixSum.begin()+ii, largePrefixSum.begin()+small.size()+ii);

	    for (size_t i = 0 ; i < outBegin.size(); ++i) {
		double normFactor = computeNormFactor(smallPrefixSum, largePrefixSum,
						      smallPrefixSum.begin(), smallPrefixSum.end()-i,
						      largePrefixSum.begin()+i+ii, largePrefixSum.begin()+ii+small.size());
	
		if (outBegin[maxSampleBegin].real()/maxNormFactorBegin < outBegin[i].real()/normFactor) {
		    maxSampleBegin = i;
		    maxNormFactorBegin = normFactor;
		}
	    }
	    //std::cout << "TEST2" << std::endl;
	    size_t maxSampleEnd = 0;
	    double maxNormFactorEnd = computeNormFactor(smallPrefixSum, largePrefixSum,
							smallPrefixSum.begin(), smallPrefixSum.end(),
							largePrefixSum.begin()+ii, largePrefixSum.begin()+small.size()+ii);
	    for (size_t i = 0 ; i < outEnd.size(); ++i) {
		double normFactor = computeNormFactor(smallPrefixSum, largePrefixSum,
						      smallPrefixSum.begin()+i, smallPrefixSum.end(),
						      largePrefixSum.begin()+ii, largePrefixSum.begin()-i+ii+small.size());

		if (outEnd[maxSampleEnd].real()/maxNormFactorEnd < outEnd[i].real()/normFactor) {
		    maxSampleEnd = i;
		    maxNormFactorEnd = normFactor;
		}
	    }
	    maxSamplesBegin[ii/small.size()+idxAdd] = Record(maxNormFactorBegin, outBegin[maxSampleBegin].real(), small.size() - maxSampleBegin);
	    maxSamplesEnd[ii/small.size()+idxAdd] = Record(maxNormFactorEnd, outEnd[maxSampleEnd].real(), small.size() - maxSampleEnd);
	}

	if (numberOfParts != pieces) break;

    }
    std::cout << '\r' << setw(8) << 100 << "%" << std::endl;

//     // FIXME: special case.
//     // small size does not divide large size
//     // => last piece is not analysed.
//     // fix this.

    for (size_t i = 0; i < maxSamplesBegin.size()-1; ++i) {
	double val = (maxSamplesBegin[i].cv + maxSamplesEnd[i+1].cv)/(maxSamplesBegin[i].nf + maxSamplesEnd[i+1].nf);
	if (val > 0.3) { // arbitrary magic number. Seems to work well.
	    size_t length = maxSamplesBegin[i].s + maxSamplesEnd[i+1].s;
	    if (length <= small.size() && length >= THRESHHOLD*small.size()) { // length must be appropriate
		results.push_back(make_pair((i+1)*small.size()-maxSamplesBegin[i].s, val));
	    }
	}
    }
}