working with fft

samplebrain/src/brain.cpp

@@ -16,29 +16,45 @@ sample brain::load_sound(std::string filename) {
     sfinfo.format=0;
     SNDFILE* f=sf_open(filename.c_str(), SFM_READ, &sfinfo);
     sample s(sfinfo.frames);
-    sf_read_float(f, s.get_non_const_buffer(), s.get_length());
+    sf_readf_float(f, s.get_non_const_buffer(), s.get_length());
     sf_close(f);
     m_samples.push_back(s);
     return s;
 }
 
+void save_sample(const string &filename, const sample s) {
+    SF_INFO sfinfo;
+    sfinfo.format=SF_FORMAT_WAV | SF_FORMAT_FLOAT;
+    sfinfo.frames=s.get_length();
+    sfinfo.samplerate=44100;
+    sfinfo.channels=1;
+    sfinfo.sections=1;
+    sfinfo.seekable=0;
+    SNDFILE* f=sf_open(filename.c_str(), SFM_WRITE, &sfinfo);
+    if (!f) cerr<<"couldn't open "<<filename<<endl;
+    u32 written = sf_writef_float(f, s.get_buffer(), s.get_length());
+    if (written!=s.get_length()) cerr<<"error: wrote "<<written<<endl;
+    sf_close(f);
+}
+
 // rewrites whole brain
-void brain::init(u32 block_size, u32 overlap) {
+void brain::init(u32 block_size, u32 overlap, bool ditchpcm) {
     m_blocks.clear();
     m_block_size = block_size;
     m_overlap = overlap;
     for (vector<sample>::iterator i=m_samples.begin(); i!=m_samples.end(); ++i) {
-        chop_and_add(*i, block_size, overlap);
+        chop_and_add(*i, block_size, overlap, ditchpcm);
     }
 }
 
-void brain::chop_and_add(const sample &s, u32 block_size, u32 overlap) {
+void brain::chop_and_add(const sample &s, u32 block_size, u32 overlap, bool ditchpcm) {
     u32 pos=0;
     while (pos+block_size-1<s.get_length()) {
+        cerr<<pos/(float)s.get_length()*100<<endl;
         sample region;
         s.get_region(region,pos,pos+block_size-1);
-        m_blocks.push_back(brain_block("",region,44100));
-        pos += block_size-overlap;
+        m_blocks.push_back(brain_block("",region,44100,ditchpcm));
+        pos += (block_size-overlap);
     }
 }
 
@@ -65,15 +81,28 @@ u32 brain::search(const brain_block &target, float ratio) const {
 
 // take another brain and rebuild this brain from bits of that one
 // (presumably this one is made from a single sample)
-sample brain::resynth(const brain &other, float ratio){
-    sample out(m_block_size*m_blocks.size());
+void brain::resynth(const string &filename, const brain &other, float ratio){
+    sample out((m_block_size-m_overlap)*m_blocks.size());
+    out.zero();
     u32 pos = 0;
+    u32 count = 0;
     for (vector<brain_block>::iterator i=m_blocks.begin(); i!=m_blocks.end(); ++i) {
+        cerr<<count/float(m_blocks.size())*100<<endl;
         u32 index = other.search(*i,ratio);
+        cerr<<index<<endl;
         out.mix(other.get_block_pcm(index),pos);
-        pos += m_block_size-m_overlap;
+
+        if (count%1000==0) {
+            save_sample(filename,out);
         }
-    return out;
+
+        ++count;
+        pos += (m_block_size-m_overlap);
+    }
+    save_sample(filename,out);
+
+    cerr<<m_blocks.size()<<" brain blocks..."<<endl;
+
 }
 
 
@@ -100,19 +129,21 @@ bool brain::unit_test() {
     assert(b.m_samples.size()==2);
     assert(b.m_blocks.size()==12);
 
-    // replicate brain
+    // replicate brains
     brain b2;
-    b2.load_sound("test_data/100f32.wav");
-    b2.load_sound("test_data/100f32.wav");
-    assert(b2.m_samples.size()==2);
-    b2.init(10, 0);
-    b.init(10, 0);
-    assert(b2.m_blocks.size()==20);
-    assert(b.search(b2.m_blocks[0],1)==0);
-    assert(b.search(b2.m_blocks[9],1)==9);
+    b2.load_sound("test_data/up.wav");
+    brain b3;
+    b3.load_sound("test_data/up.wav");
 
-    sample r = b2.resynth(b,1);
-    assert(r.get_length()==200);
+    b2.init(512, 0);
+    b3.init(512, 0);
+    assert(b3.search(b2.m_blocks[0],1)==0);
+    assert(b3.search(b2.m_blocks[9],1)==9);
+    assert(b3.search(b2.m_blocks[19],1)==19);
+    assert(b3.search(b2.m_blocks[29],1)==29);
+
+//    sample r = b2.resynth(b,1);
+//    assert(r.get_length()==200);
 
     return true;
 }

samplebrain/src/brain.h

@@ -14,24 +14,22 @@ public:
     brain();
 
     // rewrites whole brain
-    void init(u32 block_size, u32 overlap);
+    void init(u32 block_size, u32 overlap, bool ditchpcm=false);
 
     // load, chop up and add to brain
     // todo: add tags
     sample load_sound(std::string filename);
-
-    const sample &get_block_pcm(u32 index) const;
-
     // take another brain and rebuild this brain from bits of that one
     // (presumably this one is made from a single sample)
-    sample resynth(const brain &other, float ratio);
+    void resynth(const std::string &filename, const brain &other, float ratio);
 
+    const sample &get_block_pcm(u32 index) const;
     static bool unit_test();
 
 private:
 
     u32 search(const brain_block &target, float ratio) const;
-    void chop_and_add(const sample &s, u32 block_size, u32 overlap);
+    void chop_and_add(const sample &s, u32 block_size, u32 overlap, bool ditchpcm=false);
 
     vector<brain_block> m_blocks;
     vector<sample> m_samples;

samplebrain/src/brain_block.cpp

@@ -8,37 +8,38 @@ using namespace spiralcore;
 
 FFT *brain_block::m_fftw;
 
-brain_block::brain_block(const string &filename, const sample &pcm, u32 rate) :
+static const int MFCC_FILTERS=48;
+
+void enveloper(sample &s, u32 start, u32 end) {
+    for(u32 i=0; i<start; ++i) {
+        s[i]*=i/(float)start;
+    }
+    for(u32 i=0; i<end; ++i) {
+        s[(s.get_length()-1)-i]*=i/(float)end;
+    }
+}
+
+brain_block::brain_block(const string &filename, const sample &pcm, u32 rate, bool ditchpcm) :
     m_pcm(pcm),
     m_fft(pcm.get_length()),
-    m_mfcc(13),
+    m_mfcc(MFCC_FILTERS),
     m_block_size(pcm.get_length()),
     m_rate(rate),
     m_orig_filename(filename)
 {
     init_fft(m_pcm.get_length());
 
-    for (u32 i=0; i<m_block_size; i++) {
-        // convert from float to double
-        m_fftw->m_in[i] = m_pcm[i];
-    }
+    enveloper(m_pcm,50,50);
 
-    m_fftw->raw_impulse2freq();
+    m_fftw->impulse2freq(m_pcm.get_non_const_buffer(),
+                         m_fft.get_non_const_buffer());
 
-    double *spectrum = new double[m_block_size];
+    if (m_block_size>30) m_fft.crop_to(30);
+    if (ditchpcm) m_pcm.clear();
 
-    for (u32 i=0; i<m_block_size; i++) {
-        // convert from complex to float for storage
-        m_fft[i] = m_fftw->m_spectrum[i][0];
-        // convert from complex to double for mfcc calc
-        spectrum[i] = m_fftw->m_spectrum[i][0];
-    }
-
-    for (u32 i=0; i<13; i++) {
-        m_mfcc[i] = GetCoefficient(spectrum, rate, 48, m_block_size, i);
-    }
-
-    delete[] spectrum;
+//    for (u32 i=0; i<MFCC_FILTERS; i++) {
+//        m_mfcc[i] = GetCoefficient(m_fft.get_non_const_buffer(), rate, MFCC_FILTERS, m_block_size, i);
+//    }
 }
 
 void brain_block::init_fft(u32 block_size)
@@ -54,9 +55,13 @@ double brain_block::compare(const brain_block &other, float ratio) const {
     // just mfcc
     //if (ratio==1)
     {
-        for (u32 i=0; i<13; ++i) {
-            acc+=(m_mfcc[i]-other.m_mfcc[i]) * (m_mfcc[i]-other.m_mfcc[i]);
+        for (u32 i=0; i<m_fft.get_length(); ++i) {
+            acc+=(m_fft[i]-other.m_fft[i]) * (m_fft[i]-other.m_fft[i]);
         }
+
+        //for (u32 i=0; i<MFCC_FILTERS; ++i) {
+        //    acc+=(m_mfcc[i]-other.m_mfcc[i]) * (m_mfcc[i]-other.m_mfcc[i]);
+        //}
     }
     return acc;
 }
@@ -72,7 +77,7 @@ bool brain_block::unit_test() {
 
     assert(bb.m_pcm.get_length()==data.get_length());
     assert(bb.m_fft.get_length()==data.get_length());
-    assert(bb.m_mfcc.get_length()==13);
+    assert(bb.m_mfcc.get_length()==MFCC_FILTERS);
     assert(bb.m_orig_filename==string("test"));
     assert(bb.m_rate==44100);
     assert(bb.m_block_size==data.get_length());
@@ -88,6 +93,7 @@ bool brain_block::unit_test() {
     brain_block cpy("test",data,100);
     {
     brain_block bb3("test",data2,44100);
+    cerr<<bb.compare(bb3,1)<<endl;
     assert(bb.compare(bb3,1)!=0);
     cpy=bb3;
     }

samplebrain/src/brain_block.h

@@ -11,7 +11,7 @@ namespace spiralcore {
 class brain_block {
 public:
     // runs analysis on pcm
-    brain_block(const std::string &filename, const sample &pcm, u32 rate);
+    brain_block(const std::string &filename, const sample &pcm, u32 rate, bool ditchpcm=false);
 
     // returns distance based on ratio of fft-mfcc values
     double compare(const brain_block &other, float ratio) const;

samplebrain/src/libmfcc.cpp

@@ -12,18 +12,18 @@
 /*
  * Computes the specified (mth) MFCC
  *
- * spectralData - array of doubles containing the results of FFT computation. This data is already assumed to be purely real
+ * spectralData - array of mfcc_reals containing the results of FFT computation. This data is already assumed to be purely real
  * samplingRate - the rate that the original time-series data was sampled at (i.e 44100)
  * NumFilters - the number of filters to use in the computation. Recommended value = 48
  * binSize - the size of the spectralData array, usually a power of 2
  * m - The mth MFCC coefficient to compute
  *
  */
-double GetCoefficient(double* spectralData, unsigned int samplingRate, unsigned int NumFilters, unsigned int binSize, unsigned int m)
+mfcc_real GetCoefficient(mfcc_real* spectralData, unsigned int samplingRate, unsigned int NumFilters, unsigned int binSize, unsigned int m)
 {
-    double result = 0.0f;
-    double outerSum = 0.0f;
-    double innerSum = 0.0f;
+    mfcc_real result = 0.0f;
+    mfcc_real outerSum = 0.0f;
+    mfcc_real innerSum = 0.0f;
     unsigned int k, l;
 
     // 0 <= m < L
@@ -64,9 +64,9 @@ double GetCoefficient(double* spectralData, unsigned int samplingRate, unsigned
  * Computes the Normalization Factor (Equation 6)
  * Used for internal computation only - not to be called directly
  */
-double NormalizationFactor(int NumFilters, int m)
+mfcc_real NormalizationFactor(int NumFilters, int m)
 {
-    double normalizationFactor = 0.0f;
+    mfcc_real normalizationFactor = 0.0f;
 
     if(m == 0)
     {
@@ -84,14 +84,14 @@ double NormalizationFactor(int NumFilters, int m)
  * Compute the filter parameter for the specified frequency and filter bands (Eq. 2)
  * Used for internal computation only - not the be called directly
  */
-double GetFilterParameter(unsigned int samplingRate, unsigned int binSize, unsigned int frequencyBand, unsigned int filterBand)
+mfcc_real GetFilterParameter(unsigned int samplingRate, unsigned int binSize, unsigned int frequencyBand, unsigned int filterBand)
 {
-    double filterParameter = 0.0f;
+    mfcc_real filterParameter = 0.0f;
 
-    double boundary = (frequencyBand * samplingRate) / binSize;             // k * Fs / N
-    double prevCenterFrequency = GetCenterFrequency(filterBand - 1);                // fc(l - 1) etc.
-    double thisCenterFrequency = GetCenterFrequency(filterBand);
-    double nextCenterFrequency = GetCenterFrequency(filterBand + 1);
+    mfcc_real boundary = (frequencyBand * samplingRate) / binSize;             // k * Fs / N
+    mfcc_real prevCenterFrequency = GetCenterFrequency(filterBand - 1);                // fc(l - 1) etc.
+    mfcc_real thisCenterFrequency = GetCenterFrequency(filterBand);
+    mfcc_real nextCenterFrequency = GetCenterFrequency(filterBand + 1);
 
     if(boundary >= 0 && boundary < prevCenterFrequency)
     {
@@ -119,9 +119,9 @@ double GetFilterParameter(unsigned int samplingRate, unsigned int binSize, unsig
  * Compute the band-dependent magnitude factor for the given filter band (Eq. 3)
  * Used for internal computation only - not the be called directly
  */
-double GetMagnitudeFactor(unsigned int filterBand)
+mfcc_real GetMagnitudeFactor(unsigned int filterBand)
 {
-    double magnitudeFactor = 0.0f;
+    mfcc_real magnitudeFactor = 0.0f;
 
     if(filterBand >= 1 && filterBand <= 14)
     {
@@ -141,10 +141,10 @@ double GetMagnitudeFactor(unsigned int filterBand)
  * center frequencies are equally spaced on the mel scale
  * Used for internal computation only - not the be called directly
  */
-double GetCenterFrequency(unsigned int filterBand)
+mfcc_real GetCenterFrequency(unsigned int filterBand)
 {
-    double centerFrequency = 0.0f;
-    double exponent;
+    mfcc_real centerFrequency = 0.0f;
+    mfcc_real exponent;
 
     if(filterBand == 0)
     {

samplebrain/src/libmfcc.h

@@ -8,19 +8,21 @@
 
 #pragma once
 
+typedef float mfcc_real;
+
 #define PI 3.14159265358979323846264338327
 
 // Returns the specified (mth) MFCC
-double GetCoefficient(double* spectralData, unsigned int samplingRate, unsigned int NumFilters, unsigned int binSize, unsigned int m);
+mfcc_real GetCoefficient(mfcc_real* spectralData, unsigned int samplingRate, unsigned int NumFilters, unsigned int binSize, unsigned int m);
 
 // Compute the normalization factor (For internal computation only - not to be called directly)
-double NormalizationFactor(int NumFilters, int m);
+mfcc_real NormalizationFactor(int NumFilters, int m);
 
 // Compute the filter parameter for the specified frequency and filter bands (For internal computation only - not the be called directly)
-double GetFilterParameter(unsigned int samplingRate, unsigned int binSize, unsigned int frequencyBand, unsigned int filterBand);
+mfcc_real GetFilterParameter(unsigned int samplingRate, unsigned int binSize, unsigned int frequencyBand, unsigned int filterBand);
 
 // Compute the band-dependent magnitude factor for the given filter band (For internal computation only - not the be called directly)
-double GetMagnitudeFactor(unsigned int filterBand);
+mfcc_real GetMagnitudeFactor(unsigned int filterBand);
 
 // Compute the center frequency (fc) of the specified filter band (l) (For internal computation only - not the be called directly)
-double GetCenterFrequency(unsigned int filterBand);
+mfcc_real GetCenterFrequency(unsigned int filterBand);

samplebrain/src/main.cpp

@@ -25,14 +25,35 @@
 
 using namespace std;
 
-int main(int argc, char *argv[])
-{
+void unit_test() {
     cerr<<"testing brain_block"<<endl;
     if (brain_block::unit_test()) cerr<<"passed"<<endl;
     else cerr<<"failed"<<endl;
     cerr<<"testing brain"<<endl;
     if (brain::unit_test()) cerr<<"passed"<<endl;
     else cerr<<"failed"<<endl;
+}
+
+int main(int argc, char *argv[])
+{
+    unit_test();
+
+    cerr<<"starting"<<endl;
+    brain source, target;
+    source.load_sound("../sound/source/shostakovich6.wav");
+//    source.load_sound("../sound/source/eagle.wav");
+
+    target.load_sound("../sound/source/sb-right.wav");
+    cerr<<"loaded sounds"<<endl;
+
+    u32 len=3000;
+    source.init(len,len-len);
+    target.init(len,len-len/8);
+    cerr<<"ready..."<<endl;
+
+    target.resynth("shosta-sb-right.wav",source,1);
+
+
 
     //audio_device *a = new audio_device("samplebrain",44100,2048);
 }