Chris@0: /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
Chris@0: 
Chris@0: /*
Chris@0:     Sonic Visualiser
Chris@0:     An audio file viewer and annotation editor.
Chris@0:     Centre for Digital Music, Queen Mary, University of London.
Chris@0:     This file copyright 2006 Chris Cannam.
Chris@0:     
Chris@0:     This program is free software; you can redistribute it and/or
Chris@0:     modify it under the terms of the GNU General Public License as
Chris@0:     published by the Free Software Foundation; either version 2 of the
Chris@0:     License, or (at your option) any later version.  See the file
Chris@0:     COPYING included with this distribution for more information.
Chris@0: */
Chris@0: 
Chris@0: #ifndef _INTEGER_TIME_STRETCHER_H_
Chris@0: #define _INTEGER_TIME_STRETCHER_H_
Chris@0: 
Chris@0: #include "base/Window.h"
Chris@0: #include "base/RingBuffer.h"
Chris@0: 
Chris@0: #include <fftw3.h>
Chris@0: 
Chris@0: /**
Chris@0:  * A time stretcher that slows down audio by an integer multiple of
Chris@0:  * its original duration, preserving pitch.  This uses the simple
Chris@0:  * phase vocoder technique from DAFX pp275-276, adding a block-based
Chris@0:  * stream oriented API.
Chris@0:  *
Chris@0:  * Causes significant transient smearing, but sounds good for steady
Chris@0:  * notes and is generally predictable.
Chris@0:  */
Chris@0: 
Chris@0: class IntegerTimeStretcher
Chris@0: {
Chris@0: public:
Chris@12:     IntegerTimeStretcher(float ratio,
Chris@0: 			 size_t maxProcessInputBlockSize,
Chris@0: 			 size_t inputIncrement = 64,
Chris@0: 			 size_t windowSize = 2048,
Chris@0: 			 WindowType windowType = HanningWindow);
Chris@0:     virtual ~IntegerTimeStretcher();
Chris@0: 
Chris@12:     /**
Chris@12:      * Process a block.  The input array contains the given number of
Chris@12:      * samples; the output has enough space for samples * m_ratio.
Chris@12:      */
Chris@0:     void process(float *input, float *output, size_t samples);
Chris@0: 
Chris@0:     /**
Chris@0:      * Get the hop size for input.  Smaller values may produce better
Chris@0:      * results, at a cost in processing time.  Larger values are
Chris@0:      * faster but increase the likelihood of echo-like effects.  The
Chris@0:      * default is 64, which is usually pretty good, though heavy on
Chris@0:      * processor power.
Chris@0:      */
Chris@0:     size_t getInputIncrement() const { return m_n1; }
Chris@0: 
Chris@0:     /**
Chris@0:      * Get the window size for FFT processing.  Must be larger than
Chris@0:      * the input and output increments.  The default is 2048.
Chris@0:      */
Chris@0:     size_t getWindowSize() const { return m_wlen; }
Chris@0: 
Chris@0:     /**
Chris@0:      * Get the window type.  The default is a Hanning window.
Chris@0:      */
Chris@0:     WindowType getWindowType() const { return m_window->getType(); }
Chris@0: 
Chris@12:     float getRatio() const { return m_ratio; }
Chris@0:     size_t getOutputIncrement() const { return getInputIncrement() * getRatio(); }
Chris@0:     size_t getProcessingLatency() const;
Chris@0: 
Chris@0: protected:
Chris@0:     void processBlock(float *in, float *out);
Chris@0: 
Chris@12:     float m_ratio;
Chris@0:     size_t m_n1;
Chris@0:     size_t m_n2;
Chris@0:     size_t m_wlen;
Chris@0:     Window<float> *m_window;
Chris@0: 
Chris@0:     fftwf_complex *m_time;
Chris@0:     fftwf_complex *m_freq;
Chris@0:     float *m_dbuf;
Chris@12:     float *m_prevPhase;
Chris@12:     float *m_prevAdjustedPhase;
Chris@0: 
Chris@0:     fftwf_plan m_plan;
Chris@0:     fftwf_plan m_iplan;
Chris@0:     
Chris@0:     RingBuffer<float> m_inbuf;
Chris@0:     RingBuffer<float> m_outbuf;
Chris@0:     float *m_mashbuf;
Chris@0: };
Chris@0: 
Chris@0: #endif