lbajardsilogic@0: /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
lbajardsilogic@0: 
lbajardsilogic@0: /*
lbajardsilogic@0:     Sonic Visualiser
lbajardsilogic@0:     An audio file viewer and annotation editor.
lbajardsilogic@0:     Centre for Digital Music, Queen Mary, University of London.
lbajardsilogic@0:     This file copyright 2006 Chris Cannam and QMUL.
lbajardsilogic@0:     
lbajardsilogic@0:     This program is free software; you can redistribute it and/or
lbajardsilogic@0:     modify it under the terms of the GNU General Public License as
lbajardsilogic@0:     published by the Free Software Foundation; either version 2 of the
lbajardsilogic@0:     License, or (at your option) any later version.  See the file
lbajardsilogic@0:     COPYING included with this distribution for more information.
lbajardsilogic@0: */
lbajardsilogic@0: 
lbajardsilogic@0: #include "PhaseVocoderTimeStretcher.h"
lbajardsilogic@0: 
lbajardsilogic@0: #include <iostream>
lbajardsilogic@0: #include <cassert>
lbajardsilogic@0: 
lbajardsilogic@0: #include <QMutexLocker>
lbajardsilogic@0: 
lbajardsilogic@0: //#define DEBUG_PHASE_VOCODER_TIME_STRETCHER 1
lbajardsilogic@0: 
lbajardsilogic@0: PhaseVocoderTimeStretcher::PhaseVocoderTimeStretcher(size_t sampleRate,
lbajardsilogic@0:                                                      size_t channels,
lbajardsilogic@0:                                                      float ratio,
lbajardsilogic@0:                                                      bool sharpen,
lbajardsilogic@0:                                                      size_t maxOutputBlockSize) :
lbajardsilogic@0:     m_sampleRate(sampleRate),
lbajardsilogic@0:     m_channels(channels),
lbajardsilogic@0:     m_maxOutputBlockSize(maxOutputBlockSize),
lbajardsilogic@0:     m_ratio(ratio),
lbajardsilogic@0:     m_sharpen(sharpen),
lbajardsilogic@0:     m_totalCount(0),
lbajardsilogic@0:     m_transientCount(0),
lbajardsilogic@0:     m_n2sum(0),
lbajardsilogic@0:     m_mutex(new QMutex())
lbajardsilogic@0: {
lbajardsilogic@0:     initialise();
lbajardsilogic@0: }
lbajardsilogic@0: 
lbajardsilogic@0: PhaseVocoderTimeStretcher::~PhaseVocoderTimeStretcher()
lbajardsilogic@0: {
lbajardsilogic@0:     std::cerr << "PhaseVocoderTimeStretcher::~PhaseVocoderTimeStretcher" << std::endl;
lbajardsilogic@0: 
lbajardsilogic@0:     cleanup();
lbajardsilogic@0:     
lbajardsilogic@0:     delete m_mutex;
lbajardsilogic@0: }
lbajardsilogic@0: 
lbajardsilogic@0: void
lbajardsilogic@0: PhaseVocoderTimeStretcher::initialise()
lbajardsilogic@0: {
lbajardsilogic@0:     std::cerr << "PhaseVocoderTimeStretcher::initialise" << std::endl;
lbajardsilogic@0: 
lbajardsilogic@0:     calculateParameters();
lbajardsilogic@0:         
lbajardsilogic@0:     m_analysisWindow = new Window<float>(HanningWindow, m_wlen);
lbajardsilogic@0:     m_synthesisWindow = new Window<float>(HanningWindow, m_wlen);
lbajardsilogic@0: 
lbajardsilogic@0:     m_prevPhase = new float *[m_channels];
lbajardsilogic@0:     m_prevAdjustedPhase = new float *[m_channels];
lbajardsilogic@0: 
lbajardsilogic@0:     m_prevTransientMag = (float *)fftf_malloc(sizeof(float) * (m_wlen / 2 + 1));
lbajardsilogic@0:     m_prevTransientScore = 0;
lbajardsilogic@0:     m_prevTransient = false;
lbajardsilogic@0: 
lbajardsilogic@0:     m_tempbuf = (float *)fftf_malloc(sizeof(float) * m_wlen);
lbajardsilogic@0: 
lbajardsilogic@0:     m_time = new float *[m_channels];
lbajardsilogic@0:     m_freq = new fftf_complex *[m_channels];
lbajardsilogic@0:     m_plan = new fftf_plan[m_channels];
lbajardsilogic@0:     m_iplan = new fftf_plan[m_channels];
lbajardsilogic@0: 
lbajardsilogic@0:     m_inbuf = new RingBuffer<float> *[m_channels];
lbajardsilogic@0:     m_outbuf = new RingBuffer<float> *[m_channels];
lbajardsilogic@0:     m_mashbuf = new float *[m_channels];
lbajardsilogic@0: 
lbajardsilogic@0:     m_modulationbuf = (float *)fftf_malloc(sizeof(float) * m_wlen);
lbajardsilogic@0:         
lbajardsilogic@0:     for (size_t c = 0; c < m_channels; ++c) {
lbajardsilogic@0: 
lbajardsilogic@0:         m_prevPhase[c] = (float *)fftf_malloc(sizeof(float) * (m_wlen / 2 + 1));
lbajardsilogic@0:         m_prevAdjustedPhase[c] = (float *)fftf_malloc(sizeof(float) * (m_wlen / 2 + 1));
lbajardsilogic@0: 
lbajardsilogic@0:         m_time[c] = (float *)fftf_malloc(sizeof(float) * m_wlen);
lbajardsilogic@0:         m_freq[c] = (fftf_complex *)fftf_malloc(sizeof(fftf_complex) *
lbajardsilogic@0:                                                   (m_wlen / 2 + 1));
lbajardsilogic@0:         
lbajardsilogic@0:         m_plan[c] = fftf_plan_dft_r2c_1d(m_wlen, m_time[c], m_freq[c], FFTW_ESTIMATE);
lbajardsilogic@0:         m_iplan[c] = fftf_plan_dft_c2r_1d(m_wlen, m_freq[c], m_time[c], FFTW_ESTIMATE);
lbajardsilogic@0: 
lbajardsilogic@0:         m_outbuf[c] = new RingBuffer<float>
lbajardsilogic@0:             ((m_maxOutputBlockSize + m_wlen) * 2);
lbajardsilogic@0:         m_inbuf[c] = new RingBuffer<float>
lbajardsilogic@0:             (lrintf(m_outbuf[c]->getSize() / m_ratio) + m_wlen);
lbajardsilogic@0: 
lbajardsilogic@0:         std::cerr << "making inbuf size " << m_inbuf[c]->getSize() << " (outbuf size is " << m_outbuf[c]->getSize() << ", ratio " << m_ratio << ")" << std::endl;
lbajardsilogic@0: 
lbajardsilogic@0:            
lbajardsilogic@0:         m_mashbuf[c] = (float *)fftf_malloc(sizeof(float) * m_wlen);
lbajardsilogic@0:         
lbajardsilogic@0:         for (size_t i = 0; i < m_wlen; ++i) {
lbajardsilogic@0:             m_mashbuf[c][i] = 0.0;
lbajardsilogic@0:         }
lbajardsilogic@0: 
lbajardsilogic@0:         for (size_t i = 0; i <= m_wlen/2; ++i) {
lbajardsilogic@0:             m_prevPhase[c][i] = 0.0;
lbajardsilogic@0:             m_prevAdjustedPhase[c][i] = 0.0;
lbajardsilogic@0:         }
lbajardsilogic@0:     }
lbajardsilogic@0: 
lbajardsilogic@0:     for (size_t i = 0; i < m_wlen; ++i) {
lbajardsilogic@0:         m_modulationbuf[i] = 0.0;
lbajardsilogic@0:     }
lbajardsilogic@0: 
lbajardsilogic@0:     for (size_t i = 0; i <= m_wlen/2; ++i) {
lbajardsilogic@0:         m_prevTransientMag[i] = 0.0;
lbajardsilogic@0:     }
lbajardsilogic@0: }
lbajardsilogic@0: 
lbajardsilogic@0: void
lbajardsilogic@0: PhaseVocoderTimeStretcher::calculateParameters()
lbajardsilogic@0: {
lbajardsilogic@0:     std::cerr << "PhaseVocoderTimeStretcher::calculateParameters" << std::endl;
lbajardsilogic@0: 
lbajardsilogic@0:     m_wlen = 1024;
lbajardsilogic@0: 
lbajardsilogic@0:     //!!! In transient sharpening mode, we need to pick the window
lbajardsilogic@0:     //length so as to be more or less fixed in audio duration (i.e. we
lbajardsilogic@0:     //need to exploit the sample rate)
lbajardsilogic@0: 
lbajardsilogic@0:     //!!! have to work out the relationship between wlen and transient
lbajardsilogic@0:     //threshold
lbajardsilogic@0: 
lbajardsilogic@0:     if (m_ratio < 1) {
lbajardsilogic@0:         if (m_ratio < 0.4) {
lbajardsilogic@0:             m_n1 = 1024;
lbajardsilogic@0:             m_wlen = 2048;
lbajardsilogic@0:         } else if (m_ratio < 0.8) {
lbajardsilogic@0:             m_n1 = 512;
lbajardsilogic@0:         } else {
lbajardsilogic@0:             m_n1 = 256;
lbajardsilogic@0:         }
lbajardsilogic@0:         if (shouldSharpen()) {
lbajardsilogic@0:             m_wlen = 2048;
lbajardsilogic@0:         }
lbajardsilogic@0:         m_n2 = lrintf(m_n1 * m_ratio);
lbajardsilogic@0:     } else {
lbajardsilogic@0:         if (m_ratio > 2) {
lbajardsilogic@0:             m_n2 = 512;
lbajardsilogic@0:             m_wlen = 4096; 
lbajardsilogic@0:         } else if (m_ratio > 1.6) {
lbajardsilogic@0:             m_n2 = 384;
lbajardsilogic@0:             m_wlen = 2048;
lbajardsilogic@0:         } else {
lbajardsilogic@0:             m_n2 = 256;
lbajardsilogic@0:         }
lbajardsilogic@0:         if (shouldSharpen()) {
lbajardsilogic@0:             if (m_wlen < 2048) m_wlen = 2048;
lbajardsilogic@0:         }
lbajardsilogic@0:         m_n1 = lrintf(m_n2 / m_ratio);
lbajardsilogic@0:         if (m_n1 == 0) {
lbajardsilogic@0:             m_n1 = 1;
lbajardsilogic@0:             m_n2 = lrintf(m_ratio);
lbajardsilogic@0:         }
lbajardsilogic@0:     }
lbajardsilogic@0: 
lbajardsilogic@0:     m_transientThreshold = lrintf(m_wlen / 4.5);
lbajardsilogic@0: 
lbajardsilogic@0:     m_totalCount = 0;
lbajardsilogic@0:     m_transientCount = 0;
lbajardsilogic@0:     m_n2sum = 0;
lbajardsilogic@0: 
lbajardsilogic@0: 
lbajardsilogic@0:     std::cerr << "PhaseVocoderTimeStretcher: channels = " << m_channels
lbajardsilogic@0:               << ", ratio = " << m_ratio
lbajardsilogic@0:               << ", n1 = " << m_n1 << ", n2 = " << m_n2 << ", wlen = "
lbajardsilogic@0:               << m_wlen << ", max = " << m_maxOutputBlockSize << std::endl;
lbajardsilogic@0: //              << ", outbuflen = " << m_outbuf[0]->getSize() << std::endl;
lbajardsilogic@0: }
lbajardsilogic@0: 
lbajardsilogic@0: void
lbajardsilogic@0: PhaseVocoderTimeStretcher::cleanup()
lbajardsilogic@0: {
lbajardsilogic@0:     std::cerr << "PhaseVocoderTimeStretcher::cleanup" << std::endl;
lbajardsilogic@0: 
lbajardsilogic@0:     for (size_t c = 0; c < m_channels; ++c) {
lbajardsilogic@0: 
lbajardsilogic@0:         fftf_destroy_plan(m_plan[c]);
lbajardsilogic@0:         fftf_destroy_plan(m_iplan[c]);
lbajardsilogic@0: 
lbajardsilogic@0:         fftf_free(m_time[c]);
lbajardsilogic@0:         fftf_free(m_freq[c]);
lbajardsilogic@0: 
lbajardsilogic@0:         fftf_free(m_mashbuf[c]);
lbajardsilogic@0:         fftf_free(m_prevPhase[c]);
lbajardsilogic@0:         fftf_free(m_prevAdjustedPhase[c]);
lbajardsilogic@0: 
lbajardsilogic@0:         delete m_inbuf[c];
lbajardsilogic@0:         delete m_outbuf[c];
lbajardsilogic@0:     }
lbajardsilogic@0: 
lbajardsilogic@0:     fftf_free(m_tempbuf);
lbajardsilogic@0:     fftf_free(m_modulationbuf);
lbajardsilogic@0:     fftf_free(m_prevTransientMag);
lbajardsilogic@0: 
lbajardsilogic@0:     delete[] m_prevPhase;
lbajardsilogic@0:     delete[] m_prevAdjustedPhase;
lbajardsilogic@0:     delete[] m_inbuf;
lbajardsilogic@0:     delete[] m_outbuf;
lbajardsilogic@0:     delete[] m_mashbuf;
lbajardsilogic@0:     delete[] m_time;
lbajardsilogic@0:     delete[] m_freq;
lbajardsilogic@0:     delete[] m_plan;
lbajardsilogic@0:     delete[] m_iplan;
lbajardsilogic@0: 
lbajardsilogic@0:     delete m_analysisWindow;
lbajardsilogic@0:     delete m_synthesisWindow;
lbajardsilogic@0: }	
lbajardsilogic@0: 
lbajardsilogic@0: void
lbajardsilogic@0: PhaseVocoderTimeStretcher::setRatio(float ratio)
lbajardsilogic@0: {
lbajardsilogic@0:     QMutexLocker locker(m_mutex);
lbajardsilogic@0: 
lbajardsilogic@0:     size_t formerWlen = m_wlen;
lbajardsilogic@0:     m_ratio = ratio;
lbajardsilogic@0: 
lbajardsilogic@0:     std::cerr << "PhaseVocoderTimeStretcher::setRatio: new ratio " << ratio
lbajardsilogic@0:               << std::endl;
lbajardsilogic@0: 
lbajardsilogic@0:     calculateParameters();
lbajardsilogic@0: 
lbajardsilogic@0:     if (m_wlen == formerWlen) {
lbajardsilogic@0: 
lbajardsilogic@0:         // This is the only container whose size depends on m_ratio
lbajardsilogic@0: 
lbajardsilogic@0:         RingBuffer<float> **newin = new RingBuffer<float> *[m_channels];
lbajardsilogic@0: 
lbajardsilogic@0:         size_t formerSize = m_inbuf[0]->getSize();
lbajardsilogic@0:         size_t newSize = lrintf(m_outbuf[0]->getSize() / m_ratio) + m_wlen;
lbajardsilogic@0: 
lbajardsilogic@0:         std::cerr << "resizing inbuf from " << formerSize << " to "
lbajardsilogic@0:                   << newSize << " (outbuf size is " << m_outbuf[0]->getSize() << ", ratio " << m_ratio << ")" << std::endl;
lbajardsilogic@0: 
lbajardsilogic@0:         if (formerSize != newSize) {
lbajardsilogic@0: 
lbajardsilogic@0:             size_t ready = m_inbuf[0]->getReadSpace();
lbajardsilogic@0: 
lbajardsilogic@0:             for (size_t c = 0; c < m_channels; ++c) {
lbajardsilogic@0:                 newin[c] = new RingBuffer<float>(newSize);
lbajardsilogic@0:             }
lbajardsilogic@0: 
lbajardsilogic@0:             if (ready > 0) {
lbajardsilogic@0: 
lbajardsilogic@190:                 size_t copy = MIN(ready, newSize);
lbajardsilogic@0:                 float *tmp = new float[ready];
lbajardsilogic@0: 
lbajardsilogic@0:                 for (size_t c = 0; c < m_channels; ++c) {
lbajardsilogic@0:                     m_inbuf[c]->read(tmp, ready);
lbajardsilogic@0:                     newin[c]->write(tmp + ready - copy, copy);
lbajardsilogic@0:                 }
lbajardsilogic@0:                 
lbajardsilogic@0:                 delete[] tmp;
lbajardsilogic@0:             }
lbajardsilogic@0:             
lbajardsilogic@0:             for (size_t c = 0; c < m_channels; ++c) {
lbajardsilogic@0:                 delete m_inbuf[c];
lbajardsilogic@0:             }
lbajardsilogic@0:             
lbajardsilogic@0:             delete[] m_inbuf;
lbajardsilogic@0:             m_inbuf = newin;
lbajardsilogic@0:         }
lbajardsilogic@0: 
lbajardsilogic@0:     } else {
lbajardsilogic@0:         
lbajardsilogic@0:         std::cerr << "wlen changed" << std::endl;
lbajardsilogic@0:         cleanup();
lbajardsilogic@0:         initialise();
lbajardsilogic@0:     }
lbajardsilogic@0: }
lbajardsilogic@0: 
lbajardsilogic@0: size_t
lbajardsilogic@0: PhaseVocoderTimeStretcher::getProcessingLatency() const
lbajardsilogic@0: {
lbajardsilogic@0:     return getWindowSize() - getInputIncrement();
lbajardsilogic@0: }
lbajardsilogic@0: 
lbajardsilogic@0: size_t
lbajardsilogic@0: PhaseVocoderTimeStretcher::getRequiredInputSamples() const
lbajardsilogic@0: {
lbajardsilogic@0:     QMutexLocker locker(m_mutex);
lbajardsilogic@0: 
lbajardsilogic@0:     if (m_inbuf[0]->getReadSpace() >= m_wlen) return 0;
lbajardsilogic@0:     return m_wlen - m_inbuf[0]->getReadSpace();
lbajardsilogic@0: }
lbajardsilogic@0: 
lbajardsilogic@0: void
lbajardsilogic@0: PhaseVocoderTimeStretcher::putInput(float **input, size_t samples)
lbajardsilogic@0: {
lbajardsilogic@0:     QMutexLocker locker(m_mutex);
lbajardsilogic@0: 
lbajardsilogic@0:     // We need to add samples from input to our internal buffer.  When
lbajardsilogic@0:     // we have m_windowSize samples in the buffer, we can process it,
lbajardsilogic@0:     // move the samples back by m_n1 and write the output onto our
lbajardsilogic@0:     // internal output buffer.  If we have (samples * ratio) samples
lbajardsilogic@0:     // in that, we can write m_n2 of them back to output and return
lbajardsilogic@0:     // (otherwise we have to write zeroes).
lbajardsilogic@0: 
lbajardsilogic@0:     // When we process, we write m_wlen to our fixed output buffer
lbajardsilogic@0:     // (m_mashbuf).  We then pull out the first m_n2 samples from that
lbajardsilogic@0:     // buffer, push them into the output ring buffer, and shift
lbajardsilogic@0:     // m_mashbuf left by that amount.
lbajardsilogic@0: 
lbajardsilogic@0:     // The processing latency is then m_wlen - m_n2.
lbajardsilogic@0: 
lbajardsilogic@0:     size_t consumed = 0;
lbajardsilogic@0: 
lbajardsilogic@0:     while (consumed < samples) {
lbajardsilogic@0: 
lbajardsilogic@0: 	size_t writable = m_inbuf[0]->getWriteSpace();
lbajardsilogic@190: 	writable = MIN(writable, samples - consumed);
lbajardsilogic@0: 
lbajardsilogic@0: 	if (writable == 0) {
lbajardsilogic@0: #ifdef DEBUG_PHASE_VOCODER_TIME_STRETCHER
lbajardsilogic@0: 	    std::cerr << "WARNING: PhaseVocoderTimeStretcher::putInput: writable == 0 (inbuf has " << m_inbuf[0]->getReadSpace() << " samples available for reading, space for " << m_inbuf[0]->getWriteSpace() << " more)" << std::endl;
lbajardsilogic@0: #endif
lbajardsilogic@0:             if (m_inbuf[0]->getReadSpace() < m_wlen ||
lbajardsilogic@0:                 m_outbuf[0]->getWriteSpace() < m_n2) {
lbajardsilogic@0:                 std::cerr << "WARNING: PhaseVocoderTimeStretcher::putInput: Inbuf has " << m_inbuf[0]->getReadSpace() << ", outbuf has space for " << m_outbuf[0]->getWriteSpace() << " (n2 = " << m_n2 << ", wlen = " << m_wlen << "), won't be able to process" << std::endl;
lbajardsilogic@0:                 break;
lbajardsilogic@0:             }
lbajardsilogic@0: 	} else {
lbajardsilogic@0: 
lbajardsilogic@0: #ifdef DEBUG_PHASE_VOCODER_TIME_STRETCHER
lbajardsilogic@0:             std::cerr << "writing " << writable << " from index " << consumed << " to inbuf, consumed will be " << consumed + writable << std::endl;
lbajardsilogic@0: #endif
lbajardsilogic@0: 
lbajardsilogic@0:             for (size_t c = 0; c < m_channels; ++c) {
lbajardsilogic@0:                 m_inbuf[c]->write(input[c] + consumed, writable);
lbajardsilogic@0:             }
lbajardsilogic@0:             consumed += writable;
lbajardsilogic@0:         }
lbajardsilogic@0: 
lbajardsilogic@0: 	while (m_inbuf[0]->getReadSpace() >= m_wlen &&
lbajardsilogic@0: 	       m_outbuf[0]->getWriteSpace() >= m_n2) {
lbajardsilogic@0: 
lbajardsilogic@0: 	    // We know we have at least m_wlen samples available
lbajardsilogic@0: 	    // in m_inbuf.  We need to peek m_wlen of them for
lbajardsilogic@0: 	    // processing, and then read m_n1 to advance the read
lbajardsilogic@0: 	    // pointer.
lbajardsilogic@0:             
lbajardsilogic@0:             for (size_t c = 0; c < m_channels; ++c) {
lbajardsilogic@0: 
lbajardsilogic@0:                 size_t got = m_inbuf[c]->peek(m_tempbuf, m_wlen);
lbajardsilogic@0:                 assert(got == m_wlen);
lbajardsilogic@0: 
lbajardsilogic@0:                 analyseBlock(c, m_tempbuf);
lbajardsilogic@0:             }
lbajardsilogic@0: 
lbajardsilogic@0:             bool transient = false;
lbajardsilogic@0:             if (shouldSharpen()) transient = isTransient();
lbajardsilogic@0: 
lbajardsilogic@0:             size_t n2 = m_n2;
lbajardsilogic@0: 
lbajardsilogic@0:             if (transient) {
lbajardsilogic@0:                 n2 = m_n1;
lbajardsilogic@0:             }
lbajardsilogic@0: 
lbajardsilogic@0:             ++m_totalCount;
lbajardsilogic@0:             if (transient) ++m_transientCount;
lbajardsilogic@0:             m_n2sum += n2;
lbajardsilogic@0: 
lbajardsilogic@0: //            std::cerr << "ratio for last 10: " <<last10num << "/" << (10 * m_n1) << " = " << float(last10num) / float(10 * m_n1) << " (should be " << m_ratio << ")" << std::endl;
lbajardsilogic@0:             
lbajardsilogic@0:             if (m_totalCount > 50 && m_transientCount < m_totalCount) {
lbajardsilogic@0: 
lbajardsilogic@0:                 int fixed = lrintf(m_transientCount * m_n1);
lbajardsilogic@0: 
lbajardsilogic@0:                 int idealTotal = lrintf(m_totalCount * m_n1 * m_ratio);
lbajardsilogic@0:                 int idealSquashy = idealTotal - fixed;
lbajardsilogic@0: 
lbajardsilogic@0:                 int squashyCount = m_totalCount - m_transientCount;
lbajardsilogic@0:                 
lbajardsilogic@0:                 n2 = lrintf(idealSquashy / squashyCount);
lbajardsilogic@0: 
lbajardsilogic@0: #ifdef DEBUG_PHASE_VOCODER_TIME_STRETCHER
lbajardsilogic@0:                 if (n2 != m_n2) {
lbajardsilogic@0:                     std::cerr << m_n2 << " -> " << n2 << std::endl;
lbajardsilogic@0:                 }
lbajardsilogic@0: #endif
lbajardsilogic@0:             }
lbajardsilogic@0: 
lbajardsilogic@0:             for (size_t c = 0; c < m_channels; ++c) {
lbajardsilogic@0: 
lbajardsilogic@0:                 synthesiseBlock(c, m_mashbuf[c],
lbajardsilogic@0:                                 c == 0 ? m_modulationbuf : 0,
lbajardsilogic@0:                                 m_prevTransient ? m_n1 : m_n2);
lbajardsilogic@0: 
lbajardsilogic@0: 
lbajardsilogic@0: #ifdef DEBUG_PHASE_VOCODER_TIME_STRETCHER
lbajardsilogic@0:                 std::cerr << "writing first " << m_n2 << " from mashbuf, skipping " << m_n1 << " on inbuf " << std::endl;
lbajardsilogic@0: #endif
lbajardsilogic@0:                 m_inbuf[c]->skip(m_n1);
lbajardsilogic@0: 
lbajardsilogic@0:                 for (size_t i = 0; i < n2; ++i) {
lbajardsilogic@0:                     if (m_modulationbuf[i] > 0.f) {
lbajardsilogic@0:                         m_mashbuf[c][i] /= m_modulationbuf[i];
lbajardsilogic@0:                     }
lbajardsilogic@0:                 }
lbajardsilogic@0: 
lbajardsilogic@0:                 m_outbuf[c]->write(m_mashbuf[c], n2);
lbajardsilogic@0: 
lbajardsilogic@0:                 for (size_t i = 0; i < m_wlen - n2; ++i) {
lbajardsilogic@0:                     m_mashbuf[c][i] = m_mashbuf[c][i + n2];
lbajardsilogic@0:                 }
lbajardsilogic@0: 
lbajardsilogic@0:                 for (size_t i = m_wlen - n2; i < m_wlen; ++i) {
lbajardsilogic@0:                     m_mashbuf[c][i] = 0.0f;
lbajardsilogic@0:                 }
lbajardsilogic@0:             }
lbajardsilogic@0: 
lbajardsilogic@0:             m_prevTransient = transient;
lbajardsilogic@0: 
lbajardsilogic@0:             for (size_t i = 0; i < m_wlen - n2; ++i) {
lbajardsilogic@0:                 m_modulationbuf[i] = m_modulationbuf[i + n2];
lbajardsilogic@0: 	    }
lbajardsilogic@0: 
lbajardsilogic@0: 	    for (size_t i = m_wlen - n2; i < m_wlen; ++i) {
lbajardsilogic@0:                 m_modulationbuf[i] = 0.0f;
lbajardsilogic@0: 	    }
lbajardsilogic@0: 
lbajardsilogic@0:             if (!transient) m_n2 = n2;
lbajardsilogic@0: 	}
lbajardsilogic@0: 
lbajardsilogic@0: 
lbajardsilogic@0: #ifdef DEBUG_PHASE_VOCODER_TIME_STRETCHER
lbajardsilogic@0: 	std::cerr << "loop ended: inbuf read space " << m_inbuf[0]->getReadSpace() << ", outbuf write space " << m_outbuf[0]->getWriteSpace() << std::endl;
lbajardsilogic@0: #endif
lbajardsilogic@0:     }
lbajardsilogic@0: 
lbajardsilogic@0: #ifdef DEBUG_PHASE_VOCODER_TIME_STRETCHER
lbajardsilogic@0:     std::cerr << "PhaseVocoderTimeStretcher::putInput returning" << std::endl;
lbajardsilogic@0: #endif
lbajardsilogic@0: 
lbajardsilogic@0: //    std::cerr << "ratio: nominal: " << getRatio() << " actual: "
lbajardsilogic@0: //              << m_total2 << "/" << m_total1 << " = " << float(m_total2) / float(m_total1) << " ideal: " << m_ratio << std::endl;
lbajardsilogic@0: }
lbajardsilogic@0: 
lbajardsilogic@0: size_t
lbajardsilogic@0: PhaseVocoderTimeStretcher::getAvailableOutputSamples() const
lbajardsilogic@0: {
lbajardsilogic@0:     QMutexLocker locker(m_mutex);
lbajardsilogic@0: 
lbajardsilogic@0:     return m_outbuf[0]->getReadSpace();
lbajardsilogic@0: }
lbajardsilogic@0: 
lbajardsilogic@0: void
lbajardsilogic@0: PhaseVocoderTimeStretcher::getOutput(float **output, size_t samples)
lbajardsilogic@0: {
lbajardsilogic@0:     QMutexLocker locker(m_mutex);
lbajardsilogic@0: 
lbajardsilogic@0:     if (m_outbuf[0]->getReadSpace() < samples) {
lbajardsilogic@0: 	std::cerr << "WARNING: PhaseVocoderTimeStretcher::getOutput: not enough data (yet?) (" << m_outbuf[0]->getReadSpace() << " < " << samples << ")" << std::endl;
lbajardsilogic@0: 	size_t fill = samples - m_outbuf[0]->getReadSpace();
lbajardsilogic@0:         for (size_t c = 0; c < m_channels; ++c) {
lbajardsilogic@0:             for (size_t i = 0; i < fill; ++i) {
lbajardsilogic@0:                 output[c][i] = 0.0;
lbajardsilogic@0:             }
lbajardsilogic@0:             m_outbuf[c]->read(output[c] + fill, m_outbuf[c]->getReadSpace());
lbajardsilogic@0:         }
lbajardsilogic@0:     } else {
lbajardsilogic@0: #ifdef DEBUG_PHASE_VOCODER_TIME_STRETCHER
lbajardsilogic@0: 	std::cerr << "enough data - writing " << samples << " from outbuf" << std::endl;
lbajardsilogic@0: #endif
lbajardsilogic@0:         for (size_t c = 0; c < m_channels; ++c) {
lbajardsilogic@0:             m_outbuf[c]->read(output[c], samples);
lbajardsilogic@0:         }
lbajardsilogic@0:     }
lbajardsilogic@0: 
lbajardsilogic@0: #ifdef DEBUG_PHASE_VOCODER_TIME_STRETCHER
lbajardsilogic@0:     std::cerr << "PhaseVocoderTimeStretcher::getOutput returning" << std::endl;
lbajardsilogic@0: #endif
lbajardsilogic@0: }
lbajardsilogic@0: 
lbajardsilogic@0: void
lbajardsilogic@0: PhaseVocoderTimeStretcher::analyseBlock(size_t c, float *buf)
lbajardsilogic@0: {
lbajardsilogic@0:     size_t i;
lbajardsilogic@0: 
lbajardsilogic@0:     // buf contains m_wlen samples
lbajardsilogic@0: 
lbajardsilogic@0: #ifdef DEBUG_PHASE_VOCODER_TIME_STRETCHER
lbajardsilogic@0:     std::cerr << "PhaseVocoderTimeStretcher::analyseBlock (channel " << c << ")" << std::endl;
lbajardsilogic@0: #endif
lbajardsilogic@0: 
lbajardsilogic@0:     m_analysisWindow->cut(buf);
lbajardsilogic@0: 
lbajardsilogic@0:     for (i = 0; i < m_wlen/2; ++i) {
lbajardsilogic@0: 	float temp = buf[i];
lbajardsilogic@0: 	buf[i] = buf[i + m_wlen/2];
lbajardsilogic@0: 	buf[i + m_wlen/2] = temp;
lbajardsilogic@0:     }
lbajardsilogic@0: 
lbajardsilogic@0:     for (i = 0; i < m_wlen; ++i) {
lbajardsilogic@0: 	m_time[c][i] = buf[i];
lbajardsilogic@0:     }
lbajardsilogic@0: 
lbajardsilogic@0:     fftf_execute(m_plan[c]); // m_time -> m_freq
lbajardsilogic@0: }
lbajardsilogic@0: 
lbajardsilogic@0: bool
lbajardsilogic@0: PhaseVocoderTimeStretcher::isTransient()
lbajardsilogic@0: {
lbajardsilogic@0:     int count = 0;
lbajardsilogic@0: 
lbajardsilogic@0:     for (size_t i = 0; i <= m_wlen/2; ++i) {
lbajardsilogic@0: 
lbajardsilogic@0:         float real = 0.f, imag = 0.f;
lbajardsilogic@0: 
lbajardsilogic@0:         for (size_t c = 0; c < m_channels; ++c) {
lbajardsilogic@0:             real += m_freq[c][i][0];
lbajardsilogic@0:             imag += m_freq[c][i][1];
lbajardsilogic@0:         }
lbajardsilogic@0: 
lbajardsilogic@0:         float sqrmag = (real * real + imag * imag);
lbajardsilogic@0: 
lbajardsilogic@0:         if (m_prevTransientMag[i] > 0.f) {
lbajardsilogic@0:             float diff = 10.f * log10f(sqrmag / m_prevTransientMag[i]);
lbajardsilogic@0:             if (diff > 3.f) ++count;
lbajardsilogic@0:         }
lbajardsilogic@0: 
lbajardsilogic@0:         m_prevTransientMag[i] = sqrmag;
lbajardsilogic@0:     }
lbajardsilogic@0: 
lbajardsilogic@0:     bool isTransient = false;
lbajardsilogic@0: 
lbajardsilogic@0: //    if (count > m_transientThreshold &&
lbajardsilogic@0: //        count > m_prevTransientScore * 1.2) {
lbajardsilogic@0:     if (count > m_prevTransientScore &&
lbajardsilogic@0:         count > m_transientThreshold &&
lbajardsilogic@0:         count - m_prevTransientScore > int(m_wlen) / 20) {
lbajardsilogic@0:         isTransient = true;
lbajardsilogic@0: 
lbajardsilogic@0: 
lbajardsilogic@0:         std::cerr << "isTransient (count = " << count << ", prev = " << m_prevTransientScore << ", diff = " << count - m_prevTransientScore << ", ratio = " << (m_totalCount > 0 ? (float (m_n2sum) / float(m_totalCount * m_n1)) : 1.f) << ", ideal = " << m_ratio << ")" << std::endl;
lbajardsilogic@0: //    } else {
lbajardsilogic@0: //        std::cerr << " !transient (count = " << count << ", prev = " << m_prevTransientScore << ", diff = " << count - m_prevTransientScore << ")" << std::endl;
lbajardsilogic@0:     }
lbajardsilogic@0: 
lbajardsilogic@0:     m_prevTransientScore = count;
lbajardsilogic@0: 
lbajardsilogic@0:     return isTransient;
lbajardsilogic@0: }
lbajardsilogic@0: 
lbajardsilogic@0: void
lbajardsilogic@0: PhaseVocoderTimeStretcher::synthesiseBlock(size_t c,
lbajardsilogic@0:                                            float *out,
lbajardsilogic@0:                                            float *modulation,
lbajardsilogic@0:                                            size_t lastStep)
lbajardsilogic@0: {
lbajardsilogic@0:     bool unchanged = (lastStep == m_n1);
lbajardsilogic@0: 
lbajardsilogic@0:     for (size_t i = 0; i <= m_wlen/2; ++i) {
lbajardsilogic@0: 		
lbajardsilogic@0:         float phase = princargf(atan2f(m_freq[c][i][1], m_freq[c][i][0]));
lbajardsilogic@0:         float adjustedPhase = phase;
lbajardsilogic@0: 
lbajardsilogic@0:         if (!unchanged) {
lbajardsilogic@0: 
lbajardsilogic@0:             float omega = (2 * M_PI * m_n1 * i) / m_wlen;
lbajardsilogic@0: 	
lbajardsilogic@0:             float expectedPhase = m_prevPhase[c][i] + omega;
lbajardsilogic@0: 
lbajardsilogic@0:             float phaseError = princargf(phase - expectedPhase);
lbajardsilogic@0: 
lbajardsilogic@0:             float phaseIncrement = (omega + phaseError) / m_n1;
lbajardsilogic@0:             
lbajardsilogic@0:             adjustedPhase = m_prevAdjustedPhase[c][i] +
lbajardsilogic@0:                 lastStep * phaseIncrement;
lbajardsilogic@0:             
lbajardsilogic@0:             float mag = sqrtf(m_freq[c][i][0] * m_freq[c][i][0] +
lbajardsilogic@0:                               m_freq[c][i][1] * m_freq[c][i][1]);
lbajardsilogic@0:             
lbajardsilogic@0:             float real = mag * cosf(adjustedPhase);
lbajardsilogic@0:             float imag = mag * sinf(adjustedPhase);
lbajardsilogic@0:             m_freq[c][i][0] = real;
lbajardsilogic@0:             m_freq[c][i][1] = imag;
lbajardsilogic@0:         }
lbajardsilogic@0: 
lbajardsilogic@0:         m_prevPhase[c][i] = phase;
lbajardsilogic@0:         m_prevAdjustedPhase[c][i] = adjustedPhase;
lbajardsilogic@0:     }
lbajardsilogic@0: 
lbajardsilogic@0:     fftf_execute(m_iplan[c]); // m_freq -> m_time, inverse fft
lbajardsilogic@0: 
lbajardsilogic@0:     for (size_t i = 0; i < m_wlen/2; ++i) {
lbajardsilogic@0:         float temp = m_time[c][i];
lbajardsilogic@0:         m_time[c][i] = m_time[c][i + m_wlen/2];
lbajardsilogic@0:         m_time[c][i + m_wlen/2] = temp;
lbajardsilogic@0:     }
lbajardsilogic@0:     
lbajardsilogic@0:     for (size_t i = 0; i < m_wlen; ++i) {
lbajardsilogic@0:         m_time[c][i] = m_time[c][i] / m_wlen;
lbajardsilogic@0:     }
lbajardsilogic@0: 
lbajardsilogic@0:     m_synthesisWindow->cut(m_time[c]);
lbajardsilogic@0: 
lbajardsilogic@0:     for (size_t i = 0; i < m_wlen; ++i) {
lbajardsilogic@0:         out[i] += m_time[c][i];
lbajardsilogic@0:     }
lbajardsilogic@0: 
lbajardsilogic@0:     if (modulation) {
lbajardsilogic@0: 
lbajardsilogic@0:         float area = m_analysisWindow->getArea();
lbajardsilogic@0: 
lbajardsilogic@0:         for (size_t i = 0; i < m_wlen; ++i) {
lbajardsilogic@0:             float val = m_synthesisWindow->getValue(i);
lbajardsilogic@0:             modulation[i] += val * area;
lbajardsilogic@0:         }
lbajardsilogic@0:     }
lbajardsilogic@0: }
lbajardsilogic@0: 
lbajardsilogic@0: