Merge pull request #550 from pabloEntropia/gfcc-thresholding

GFCC: Add silenceThreshold parameter #543
MTG · Jan 5, 2017 · fe011a0 · fe011a0
2 parents 9a35819 + 3d354ff
commit fe011a0
Show file tree

Hide file tree

Showing 3 changed files with 27 additions and 29 deletions.
diff --git a/src/algorithms/spectral/gfcc.cpp b/src/algorithms/spectral/gfcc.cpp
@@ -45,7 +45,11 @@ void GFCC::configure() {
                   "outputSize", parameter("numberCoefficients"),
                   INHERIT("dctType"));
   _logbands.resize(parameter("numberBands").toInt());
-  setCompressor(parameter("logType").toString());
+
+  _logType = parameter("logType").toLower();
+  _silenceThreshold = parameter("silenceThreshold").toReal();
+  _dbSilenceThreshold = 10 * log10(_silenceThreshold);
+  _logSilenceThreshold = log(_silenceThreshold);
 }
 
 void GFCC::compute() {
@@ -59,11 +63,22 @@ void GFCC::compute() {
   _gtFilter->output("bands").set(bands);
   _gtFilter->compute();
 
-
-
-
   for (int i=0; i<int(bands.size()); ++i) {
-    _logbands[i] = (*_compressor)(bands[i]);
+    if (_logType == "dbpow") {
+       _logbands[i] = pow2db(bands[i], _silenceThreshold, _dbSilenceThreshold);
+     }
+     else if (_logType == "dbamp") {
+       _logbands[i] = amp2db(bands[i], _silenceThreshold, _dbSilenceThreshold);
+     }
+     else if (_logType == "log") {
+      _logbands[i] = lin2log(bands[i], _silenceThreshold, _logSilenceThreshold);
+     }
+     else if (_logType == "natural") {
+       _logbands[i] = bands[i];
+     }
+     else {
+       throw EssentiaException("GFCC: Bad 'logType' parameter");
+     }
   }
 
   // compute the DCT of these bands
@@ -72,21 +87,3 @@ void GFCC::compute() {
   _dct->compute();
 }
 
-void GFCC::setCompressor(std::string logType){
-  if (logType == "natural"){
-    _compressor = linear;
-  }
-  else if (logType == "dbpow"){
-    _compressor = pow2db;
-  }
-  else if (logType == "dbamp"){
-    _compressor = amp2db;
-  }
-  else if (logType == "log"){
-    _compressor = log;
-  }
-  else{
-    throw EssentiaException("GFCC: Bad 'logType' parameter");
-  }
-
-}
diff --git a/src/algorithms/spectral/gfcc.h b/src/algorithms/spectral/gfcc.h
@@ -37,11 +37,10 @@ class GFCC : public Algorithm {
 
   std::vector<Real> _logbands;
 
-  typedef  Real (*funcPointer)(Real);
-  funcPointer _compressor;
-
-  void setCompressor(std::string logType);
-
+  std::string _logType;
+  Real _silenceThreshold;
+  Real _dbSilenceThreshold;
+  Real _logSilenceThreshold;
 
  public:
   GFCC() {
@@ -66,6 +65,7 @@ class GFCC : public Algorithm {
     declareParameter("lowFrequencyBound", "the lower bound of the frequency range [Hz]", "[0,inf)", 40.);
     declareParameter("highFrequencyBound", "the upper bound of the frequency range [Hz]", "(0,inf)", 22050.);
     declareParameter("type", "use magnitude or power spectrum","{magnitude,power}", "power");
+    declareParameter("silenceThreshold", "silence threshold for computing log-energy bands", "(0,inf)", 1e-9);
     declareParameter("logType","logarithmic compression type. Use 'dbpow' if working with power and 'dbamp' if working with magnitudes","{natural,dbpow,dbamp,log}","dbamp");
     declareParameter("dctType", "the DCT type", "[2,3]", 2);
   }

diff --git a/test/src/unittest/spectral/test_gfcc.py b/test/src/unittest/spectral/test_gfcc.py
@@ -30,6 +30,7 @@ def InitGFCC(self, numCoeffs, logType):
                     sampleRate = 44100,
                     numberBands = 40,
                     numberCoefficients = numCoeffs,
+                    silenceThreshold = 1e-9,
                     lowFrequencyBound = 0,
                     highFrequencyBound = 11000,
                     logType = logType)
@@ -52,7 +53,7 @@ def testRegression(self):
     def testZero(self):
         # zero input should return dct(lin2db(0)). Try with different sizes
         size = 1025
-        val = amp2db(0)
+        val = 2 * 10 * np.log10(1e-9)
         expected = DCT(inputSize=40, outputSize=13)([val for x in range(40)])
         while (size > 256 ):
             bands, gfcc = GFCC(inputSize = size)(zeros(size))