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@14: #ifndef _PHASE_VOCODER_TIME_STRETCHER_H_
Chris@14: #define _PHASE_VOCODER_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@14:  * A time stretcher that alters the performance speed of audio,
Chris@14:  * preserving pitch.  This uses the simple phase vocoder technique
Chris@14:  * from DAFX pp275-276, adding a block-based 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@14: class PhaseVocoderTimeStretcher
Chris@0: {
Chris@0: public:
Chris@15:     PhaseVocoderTimeStretcher(float ratio, size_t maxProcessInputBlockSize);
Chris@14:     virtual ~PhaseVocoderTimeStretcher();
Chris@0: 
Chris@12:     /**
Chris@12:      * Process a block.  The input array contains the given number of
Chris@15:      * samples; the output must have space for lrintf(samples * m_ratio).
Chris@12:      */
Chris@0:     void process(float *input, float *output, size_t samples);
Chris@0: 
Chris@0:     /**
Chris@15:      * Get the hop size for input.
Chris@0:      */
Chris@0:     size_t getInputIncrement() const { return m_n1; }
Chris@0: 
Chris@0:     /**
Chris@15:      * Get the hop size for output.
Chris@15:      */
Chris@15:     size_t getOutputIncrement() const { return getInputIncrement() * getRatio(); }
Chris@15: 
Chris@15:     /**
Chris@15:      * Get the window size for FFT processing.
Chris@0:      */
Chris@0:     size_t getWindowSize() const { return m_wlen; }
Chris@0: 
Chris@0:     /**
Chris@15:      * Get the window type.
Chris@0:      */
Chris@0:     WindowType getWindowType() const { return m_window->getType(); }
Chris@0: 
Chris@15:     /**
Chris@15:      * Get the stretch ratio set in the constructor.
Chris@15:      */
Chris@12:     float getRatio() const { return m_ratio; }
Chris@15: 
Chris@15:     /**
Chris@15:      * Get the latency added by the time stretcher, in sample frames.
Chris@15:      */
Chris@0:     size_t getProcessingLatency() const;
Chris@0: 
Chris@0: protected:
Chris@13:     /**
Chris@13:      * Process a single phase vocoder frame.
Chris@13:      * 
Chris@13:      * Take m_wlen time-domain source samples from in, perform an FFT,
Chris@13:      * phase shift, and IFFT, and add the results to out (presumably
Chris@13:      * overlapping parts of existing data from prior frames).
Chris@13:      *
Chris@13:      * Also add to the modulation output the results of windowing a
Chris@13:      * set of 1s with the resynthesis window -- this can then be used
Chris@13:      * to ensure the output has the correct magnitude in cases where
Chris@13:      * the window overlap varies or otherwise results in something
Chris@13:      * other than a flat sum.
Chris@13:      */
Chris@13:     void processBlock(float *in, float *out, float *modulation);
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@15:     RingBuffer<float> *m_inbuf;
Chris@15:     RingBuffer<float> *m_outbuf;
Chris@0:     float *m_mashbuf;
Chris@13:     float *m_modulationbuf;
Chris@0: };
Chris@0: 
Chris@0: #endif