MIR::anonymous_namespace{MirDsp.cpp} Namespace Reference

Functions
float	GetNoveltyMeasure (const PffftFloatVector &prevPowSpec, const PffftFloatVector &powSpec)
std::vector< float >	GetMovingAverage (const std::vector< float > &x, double hopRate)

Function Documentation

GetMovingAverage()

std::vector< float > MIR::anonymous_namespace{MirDsp.cpp}::GetMovingAverage	(	const std::vector< float > &	x,
		double	hopRate
	)

Definition at line 39 of file MirDsp.cpp.

{
   constexpr auto smoothingWindowDuration = 0.2;
   // An odd number.
   const int M = std::round(smoothingWindowDuration * hopRate / 4) * 2 + 1;
   const auto window = GetNormalizedHann(2 * M + 1);
   auto n = 0;
   std::vector<float> movingAverage(x.size());
   std::transform(x.begin(), x.end(), movingAverage.begin(), [&](float) {
      const auto m = IotaRange(-M, M + 1);
      const auto y =
         std::accumulate(m.begin(), m.end(), 0.f, [&](float y, int i) {
            auto k = n + i;
            while (k < 0)
               k += x.size();
            while (k >= x.size())
               k -= x.size();
            return y + x[k] * window[i + M];
         });
      ++n;
      // The moving average of the raw ODF will be subtracted from it to yield
      // the final ODF, negative results being set to 0. (This is to remove
      // noise of small ODF peaks before the method's quantization step.) The
      // larger this multiplier, the less peaks will remain. This value was
      // found by trial and error, using the benchmarking framework
      // (see TatumQuantizationFitBenchmarking.cpp)
      constexpr auto thresholdRaiser = 1.5f;
      return y * thresholdRaiser;
   });
   return movingAverage;
}

References MIR::GetNormalizedHann(), and fast_float::round().

Referenced by MIR::GetOnsetDetectionFunction().

Here is the call graph for this function:

Here is the caller graph for this function:

GetNoveltyMeasure()

float MIR::anonymous_namespace{MirDsp.cpp}::GetNoveltyMeasure	(	const PffftFloatVector &	prevPowSpec,
		const PffftFloatVector &	powSpec
	)

Definition at line 28 of file MirDsp.cpp.

{
   auto k = 0;
   return std::accumulate(
      powSpec.begin(), powSpec.end(), 0.f, [&](float a, float mag) {
         // Half-wave-rectified stuff
         return a + std::max(0.f, mag - prevPowSpec[k++]);
      });
}

Referenced by MIR::GetOnsetDetectionFunction().

Here is the caller graph for this function: