WSOLA. More...

#include <Nsound/Stretcher.h>

Public Member Functions
	Stretcher (const float64 &sample_rate, const float64 &window_size_seconds=0.08, const float64 &max_delta_window=0.25)
	Default Constructor. More...

	Stretcher (const Stretcher &copy)
	Copy constructor. More...

virtual	~Stretcher ()
	Destructor. More...

Stretcher &	operator= (const Stretcher &rhs)
	Assignment. More...

AudioStream	pitchShift (const AudioStream &x, const float64 &factor)

AudioStream	pitchShift (const AudioStream &x, const Buffer &factor)

Buffer	pitchShift (const Buffer &x, const float64 &factor)

Buffer	pitchShift (const Buffer &x, const Buffer &factor)

void	showProgress (boolean flag)

AudioStream	timeShift (const AudioStream &x, const float64 &factor)

AudioStream	timeShift (const AudioStream &x, const Buffer &factor)

Buffer	timeShift (const Buffer &x, const float64 &factor)

Buffer	timeShift (const Buffer &x, const Buffer &factor)

Protected Member Functions
void	analyize (const Buffer &input, const float64 &factor)

void	analyize (const Buffer &input, const Buffer &factor)

uint32	searchForBestMatch (const Buffer &input, uint32 source_index, uint32 search_index) const

Buffer	overlapAdd (const Buffer &input) const

Protected Attributes
Buffer *	frames_

float64	sample_rate_

Buffer *	window_

uint32	window_length_

uint32	max_delta_

boolean	show_progress_

Detailed Description

WSOLA.

Definition at line 49 of file Stretcher.h.

Constructor & Destructor Documentation

Stretcher::Stretcher	(	const float64 &	sample_rate,
		const float64 &	window_size_seconds = `0.08`,
		const float64 &	max_delta_window = `0.25`
	)

Default Constructor.

sample_rate: the sample rate window_size: default window size in seconds max_delta_window: percent of window size to search for best fit

Definition at line 54 of file Stretcher.cc.

References Nsound::Generator::drawWindowHanning(), frames_, Nsound::Buffer::getLength(), max_delta_, window_, and window_length_.

     :
     frames_(NULL),
     sample_rate_(sample_rate),
     window_(NULL),
     window_length_(0),
     max_delta_(0),
     show_progress_(false)
 {
     frames_ = new Buffer(1024);
     window_ = new Buffer(1024);
 
     Sine sin(sample_rate);
     *window_ = sin.drawWindowHanning(window_size_seconds);
     window_length_ = window_->getLength();
     max_delta_ = uint32(float64(window_length_) * max_delta_window);
 }

Stretcher::Stretcher ( const Stretcher & copy )

Copy constructor.

Definition at line 77 of file Stretcher.cc.

References frames_, window_, and window_length_.

     :
     frames_(NULL),
     sample_rate_(copy.sample_rate_),
     window_(NULL),
     window_length_(copy.window_length_),
     max_delta_(copy.max_delta_),
     show_progress_(copy.show_progress_)
 {
     frames_ = new Buffer(1024);
     window_ = new Buffer(window_length_);
 
     *this = copy;
 }

Stretcher::~Stretcher ( )

virtual

Destructor.

Definition at line 94 of file Stretcher.cc.

References frames_, and window_.

 {
     delete frames_;
     delete window_;
 }

Member Function Documentation

Stretcher & Stretcher::operator= ( const Stretcher & rhs )

Assignment.

Definition at line 103 of file Stretcher.cc.

References frames_, max_delta_, sample_rate_, show_progress_, window_, and window_length_.

 {
     if(this == & rhs) return *this;
 
     sample_rate_   = rhs.sample_rate_;
     window_length_ = rhs.window_length_;
     max_delta_     = rhs.max_delta_;
     *frames_       = *rhs.frames_;
     *window_       = *rhs.window_;
     show_progress_ = rhs.show_progress_;
 
     return *this;
 }

AudioStream Stretcher::pitchShift	(	const AudioStream &	x,
		const float64 &	factor
	)

Definition at line 338 of file Stretcher.cc.

References analyize(), Nsound::AudioStream::getMono(), Nsound::AudioStream::getNChannels(), Nsound::AudioStream::getSampleRate(), and overlapAdd().

Referenced by main().

 {
     // Prepare for time shift.
     analyize(x.getMono()[0], factor);
 
     AudioStream y(x.getSampleRate(), x.getNChannels());
 
     for(uint32 i = 0; i < x.getNChannels(); ++i)
     {
         y[i] = overlapAdd(x[i]);
     }
 
     return y.getResample(1.0/factor);
 }

AudioStream Stretcher::pitchShift	(	const AudioStream &	x,
		const Buffer &	factor
	)

Definition at line 355 of file Stretcher.cc.

References analyize(), Nsound::AudioStream::getMono(), Nsound::AudioStream::getNChannels(), Nsound::AudioStream::getSampleRate(), and overlapAdd().

 {
     // Prepare for time shift.
     analyize(x.getMono()[0], factor);
 
     AudioStream y(x.getSampleRate(), x.getNChannels());
 
     for(uint32 i = 0; i < x.getNChannels(); ++i)
     {
         y[i] = overlapAdd(x[i]);
     }
 
     return y.getSpeedUp(factor);
 }

Buffer Stretcher::pitchShift	(	const Buffer &	x,
		const float64 &	factor
	)

Definition at line 372 of file Stretcher.cc.

References analyize(), Nsound::Buffer::getResample(), and overlapAdd().

 {
     // Prepare for time shift.
     analyize(x, factor);
 
     Buffer stretched = overlapAdd(x);
 
     return stretched.getResample(1.0/factor);
 }

Buffer Stretcher::pitchShift	(	const Buffer &	x,
		const Buffer &	factor
	)

Definition at line 384 of file Stretcher.cc.

References analyize(), Nsound::Buffer::getSpeedUp(), and overlapAdd().

 {
     // Prepare for time shift.
     analyize(x, factor);
 
     Buffer stretched = overlapAdd(x);
 
     return stretched.getSpeedUp(factor);
 }

void Nsound::Stretcher::showProgress ( boolean flag )

inline

Definition at line 86 of file Stretcher.h.

References show_progress_.

Referenced by main(), and stretch_to().

86 {show_progress_ = flag;};

Nsound::Stretcher::show_progress_

boolean show_progress_

Definition: Stretcher.h:122

AudioStream Stretcher::timeShift	(	const AudioStream &	x,
		const float64 &	factor
	)

Definition at line 396 of file Stretcher.cc.

References analyize(), Nsound::AudioStream::getMono(), Nsound::AudioStream::getNChannels(), Nsound::AudioStream::getSampleRate(), and overlapAdd().

Referenced by main().

 {
     // Prepare for time shift.
     analyize(x.getMono()[0], factor);
 
     AudioStream y(x.getSampleRate(), x.getNChannels());
 
     for(uint32 i = 0; i < x.getNChannels(); ++i)
     {
         y[i] = overlapAdd(x[i]);
     }
 
     return y;
 }

AudioStream Stretcher::timeShift	(	const AudioStream &	x,
		const Buffer &	factor
	)

Definition at line 413 of file Stretcher.cc.

References analyize(), Nsound::AudioStream::getMono(), Nsound::AudioStream::getNChannels(), Nsound::AudioStream::getSampleRate(), and overlapAdd().

 {
     // Prepare for time shift.
     analyize(x.getMono()[0], factor);
 
     AudioStream y(x.getSampleRate(), x.getNChannels());
 
     for(uint32 i = 0; i < x.getNChannels(); ++i)
     {
         y[i] = overlapAdd(x[i]);
     }
 
     return y;
 }

Buffer Stretcher::timeShift	(	const Buffer &	x,
		const float64 &	factor
	)

Definition at line 430 of file Stretcher.cc.

References analyize(), and overlapAdd().

 {
     // Prepare for time shift.
     analyize(x, factor);
 
     Buffer stretched = overlapAdd(x);
 
     return stretched;
 }

Buffer Stretcher::timeShift	(	const Buffer &	x,
		const Buffer &	factor
	)

Definition at line 442 of file Stretcher.cc.

References analyize(), and overlapAdd().

 {
     // Prepare for time shift.
     analyize(x, factor);
 
     Buffer stretched = overlapAdd(x);
 
     return stretched;
 }

void Stretcher::analyize	(	const Buffer &	input,
		const float64 &	factor
	)

protected

Definition at line 119 of file Stretcher.cc.

Referenced by pitchShift(), and timeShift().

 {
     Buffer factor(1);
     factor << f;
 
     analyize(input, factor);
 }

void Stretcher::analyize	(	const Buffer &	input,
		const Buffer &	factor
	)

protected

Definition at line 129 of file Stretcher.cc.

References Nsound::Buffer::cbegin(), frames_, Nsound::Buffer::getLength(), searchForBestMatch(), show_progress_, and window_length_.

 {
     // Always calculate time shift.
 
     Buffer::const_circular_iterator ifactor = factor.cbegin();
 
     uint32 tau = static_cast<uint32>(window_length_ / 2.0 / ::fabs(*ifactor));
 
     uint32 input_length = input.getLength();
 
     *frames_ = Buffer(256);
 
     uint32 i = 0;
     while(i < input_length - window_length_)
     {
         *frames_ << i;
         i += tau;
 
         ifactor += tau;
 
         tau = static_cast<uint32>(window_length_ / 2.0 / ::fabs(*ifactor));
     }
 
     uint32 n_frames = frames_->getLength();
 
     uint32 mod_frames = n_frames / 100;
 
     if(n_frames < 100) mod_frames = n_frames;
 
     if(show_progress_)
     {
         #ifdef NSOUND_CUDA
 
             printf("Using CUDA\n");
 
         #else
 
             #ifdef NSOUND_OPENMP
                 #pragma omp parallel
                 {
                     if(0 == omp_get_thread_num())
                     {
                         printf("Using OpenMP with %d threads\n",
                             omp_get_num_threads());
                         fflush(stdout);
                     }
                 }
             #endif
 
         #endif
 
         printf("Analyizing %3.0f%%", 0.0);
         fflush(stdout);
     }
 
     for(uint32 j = 0; j < n_frames-1; ++j)
     {
         if(show_progress_ && j % mod_frames == 0)
         {
             printf("\b\b\b\b");
             printf("%3.0f%%", 100.0 * float64(j) / float64(n_frames));
             fflush(stdout);
         }
 
         (*frames_)[j] = searchForBestMatch(
             input,
             uint32((*frames_)[j]),
             uint32((*frames_)[j+1]));
     }
 
     if(show_progress_)
     {
         printf("\b\b\b\b");
         printf("%3.0f%%\n", 100.0);
         fflush(stdout);
     }
 }

uint32 Stretcher::searchForBestMatch	(	const Buffer &	input,
		uint32	source_index,
		uint32	search_index
	)		const

protected

Definition at line 209 of file Stretcher.cc.

References Nsound::FFTransform::fft(), Nsound::Buffer::getSum(), max_delta_, Nsound::Buffer::subbuffer(), and window_length_.

Referenced by analyize().

 {
     // Nick's original score function
     #if 1
         Buffer b_source = input.subbuffer(source_index, window_length_);
 
         uint32 t = 0;
         float64 min_score = 1.0e100;
 
         #ifdef NSOUND_OPENMP
             #pragma omp parallel for shared(t, min_score)
         #endif
         for(uint32 i = 0; i < max_delta_; ++i)
         {
             Buffer b_test = input.subbuffer(search_index + i, window_length_)
                           - b_source;
 
             b_test *= b_test;
 
             float64 score = b_test.getSum();
 
             if(score < min_score)
             {
                 min_score = score;
                 t = i;
             }
         }
     #endif
 
     #if 0
         // Maximizing fft phase information.
 
         FFTransform transform(100.0); // Sample rate here won't be used.
 
         Buffer b_source = input.subbuffer(source_index, window_length_);
 
         FFTChunkVector vec = transform.fft(b_source, window_length_);
 
         Buffer source_fft = vec[0].getPhase();
 
         uint32 t = 0;
         float64 min_score = 1.0e100;
         for(uint32 i = 0; i < max_delta_; ++i)
         {
             Buffer b_test = input.subbuffer(search_index + i, window_length_);
 
             FFTChunkVector vec = transform.fft(b_test, window_length_);
 
             Buffer test_fft = vec[0].getPhase();
 
             Buffer delta = source_fft - test_fft;
 
             delta *= delta;
 
             float64 score = delta.getSum();
 
             if(score < min_score)
             {
                 min_score = score;
                 t = i;
             }
         }
     #endif
 
     return t + search_index;
 }

Buffer Stretcher::overlapAdd ( const Buffer & input ) const

protected

Definition at line 281 of file Stretcher.cc.

References Nsound::Buffer::add(), Nsound::Generator::drawLine(), frames_, Nsound::Buffer::getLength(), Nsound::Buffer::getReverse(), sample_rate_, show_progress_, Nsound::Buffer::subbuffer(), window_, and window_length_.

Referenced by pitchShift(), and timeShift().

 {
     Sine sin(sample_rate_);
 
     float64 half_window_seconds = (window_length_ / 2) / sample_rate_;
 
     uint32 half_length = window_length_ / 2;
 
     Buffer window_start = sin.drawLine(half_window_seconds, 1.0, 1.0)
                        << window_->subbuffer(half_length, half_length);
 
     Buffer window_finish = window_start.getReverse();
 
     Buffer output = input.subbuffer(0, window_length_) * (*window_);
 
     uint32 n_frames = frames_->getLength();
 
     uint32 mod_frames = n_frames / 100;
 
     if(n_frames < 100) mod_frames = n_frames;
 
     if(show_progress_)
     {
         printf("Creating output %3.0f%%", 0.0);
         fflush(stdout);
     }
 
     for(uint32 i = 1; i < n_frames; ++i)
     {
         if(show_progress_ && i % mod_frames == 0)
         {
             printf("\b\b\b\b");
             printf("%3.0f%%", 100.0 * float64(i) / float64(n_frames));
             fflush(stdout);
         }
 
         uint32 frame = uint32((*frames_)[i]);
 
         Buffer next_window = input.subbuffer(frame, window_length_)
                            * (*window_);
 
         // overlap & add
         output.add(next_window, i * half_length);
     }
 
     if(show_progress_)
     {
         printf("\b\b\b\b");
         printf("%3.0f%%\n", 100.0);
         fflush(stdout);
     }
 
     return output;
 }