Chris@0: /*
Chris@0:   copyright (C) 2011 I. Irigaray, M. Rocamora
Chris@0: 
Chris@0:   This program is free software: you can redistribute it and/or modify
Chris@0:   it under the terms of the GNU General Public License as published by
Chris@0:   the Free Software Foundation, either version 3 of the License, or
Chris@0:   (at your option) any later version.
Chris@0: 
Chris@0:   This program is distributed in the hope that it will be useful,
Chris@0:   but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@0:   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
Chris@0:   GNU General Public License for more details.
Chris@0: 
Chris@0:   You should have received a copy of the GNU General Public License
Chris@0:   along with this program.  If not, see <http://www.gnu.org/licenses/>.
Chris@7: */
Chris@0: 
Chris@0: // Remember to use a different guard symbol in each header!
Chris@12: #ifndef FCHTRANSFORMF0GRAM_H
Chris@12: #define FCHTRANSFORMF0GRAM_H
Chris@7: 
Chris@0: #define _USE_MATH_DEFINES
Chris@0: #include <cmath>
Chris@0: #include <vamp-sdk/Plugin.h>
Chris@14: #include <vamp-sdk/FFT.h>
Chris@0: #include <complex>
Chris@0: #include <iostream>
Chris@0: #include <fstream>
Chris@0: #include <string.h>
Chris@0: 
Chris@0: using namespace std;
Chris@0: using std::string;
Chris@0: 
Chris@14: using _VampPlugin::Vamp::FFTReal;
Chris@14: 
Chris@15: class FChTransformF0gram : public Vamp::Plugin
Chris@15: {
Chris@0: public:
Chris@15:     enum ProcessingMode {
Chris@15:         ModeF0Gram,
Chris@15:         ModeSpectrogram,
Chris@15:         ModeRoughSpectrogram
Chris@15:     };
Chris@15:     
Chris@15:     FChTransformF0gram(ProcessingMode mode, float inputSampleRate);
Chris@0:     virtual ~FChTransformF0gram();
Chris@0: 
Chris@0:     string getIdentifier() const;
Chris@0:     string getName() const;
Chris@0:     string getDescription() const;
Chris@0:     string getMaker() const;
Chris@0:     string getCopyright() const;
Chris@0:     int getPluginVersion() const;
Chris@0: 
Chris@0:     InputDomain getInputDomain() const;
Chris@0:     size_t getMinChannelCount() const;
Chris@0:     size_t getMaxChannelCount() const;
Chris@0:     size_t getPreferredStepSize() const;
Chris@0:     size_t getPreferredBlockSize() const;
Chris@0: 
Chris@0:     ParameterList getParameterDescriptors() const;
Chris@0:     float getParameter(string identifier) const;
Chris@0:     void setParameter(string identifier, float value);
Chris@0: 
Chris@0:     ProgramList getPrograms() const;
Chris@0: 
Chris@0:     OutputList getOutputDescriptors() const;
Chris@0: 
Chris@0:     bool initialise(size_t channels, size_t stepSize, size_t blockSize);
Chris@0:     void reset();
Chris@0: 
Chris@0:     FeatureSet process(const float *const *inputBuffers,
Chris@7:                        Vamp::RealTime timestamp);
Chris@0: 
Chris@0:     FeatureSet getRemainingFeatures();
Chris@0: 
Chris@0: protected:
Chris@15:     ProcessingMode m_processingMode;
Chris@20:     bool m_initialised;
Chris@10:     int m_stepSize;
Chris@10:     int m_blockSize;
Chris@0:     float m_fs; // input sampling rate (inputSampleRate)
Chris@20:     float *m_inputBuffer;
Chris@0: 
Chris@0:     // plugin-specific data and methods go here
Chris@0: 
Chris@0:     // =============  WARPING PARAMETERS  =============
Chris@0: 
Chris@0:     double m_fmax; // maximum frequency of interest (Hz)
Chris@10:     int m_num_f0s; // number of f0 values in F0gram grid
Chris@0:     double *m_f0s; // vector of f0 values
Chris@0: 
Chris@24:     vector<int> nsamp_options;
Chris@24:     
Chris@0:     typedef struct {
Chris@10:         int nsamps_twarp; // number of samples of the warped signal frame
Chris@0:         double alpha_max; // maximum value of normalized frequency deviation (alpha)
Chris@10:         int num_warps; // number of warpings
Chris@10:         int fact_over_samp; // oversampling factor
Chris@10:         int alpha_dist; // distribution of alpha values, 'lin' or 'log' (0 - 1)
Chris@0:     } warping_parameters;
Chris@0: 
Chris@0:     warping_parameters m_warp_params;
Chris@0: 
Chris@0:     // =============   F0-GRAM PARAMETERS  =============
Chris@0: 
Chris@0:     typedef struct {
Chris@0:         double f0min; // minimun fundamental frequency
Chris@10:         int num_octs; // number of octaves
Chris@10:         int num_f0s_per_oct; // number of f0s per octave
Chris@0:         bool prefer; // whether to use a f0 preference guassian function
Chris@10:         int prefer_mean; // mean of f0 preference function (MIDI number for C4)
Chris@10:         int prefer_stdev; // stdev of f0 preference function (stdev in MIDI numbers)
Chris@0:     } f0_parameters;
Chris@0: 
Chris@0:     f0_parameters m_f0_params;
Chris@15: 
Chris@15:     enum F0GramMode {
Chris@15:         AllBinsOfBestDirection,
Chris@15:         BestBinOfAllDirections
Chris@15:     };
Chris@15:     F0GramMode m_f0gram_mode;
Chris@0: 
Chris@0:     // ======== GATHERED LOG SPECTRUM PARAMETERS =======
Chris@0: 
Chris@0:     typedef struct {
Chris@0:         bool HP_logS; //high-pass logS
Chris@0:         int att_subharms; // whether to attenuate subharmonics
Chris@0:         // model parameter variables (default values)
Chris@0:         double median_poly_coefs[3];
Chris@0:         double sigma_poly_coefs[3];
Chris@0:     } glogs_parameters;
Chris@0: 
Chris@0:     glogs_parameters m_glogs_params;
Chris@0: 
Chris@0:     // =============  WARPING DESIGN  =============
Chris@0: 
Chris@0:     typedef struct {
Chris@0:         double fs_orig; // sampling frequency after oversampling
Chris@0:         double fs_warp; // sampling frequency of warped signal
Chris@0:         double *chirp_rates; // chirp rates
Chris@10:         int nsamps_torig; // number of samples of the original signal frame
Chris@10:         int *pos_int; // index of previous sample to do the warping by interpolation efficiently
Chris@0:         double *pos_frac; // fractional value to do the warping by interpolation efficiently
Chris@0:     } warping_design;
Chris@0: 
Chris@0:     warping_design m_warpings;
Chris@0:     // LPFWindow
Chris@0:     double *mp_LPFWindow;
Chris@0:     double *LPF_time;
Chris@14:     double *LPF_frequency;
Chris@14:     FFTReal *fft_forward_LPF; // two of these as they have different sizes
Chris@14:     FFTReal *fft_inverse_LPF;
Chris@7:     // timeWindow
Chris@7:     double *m_timeWindow;
Chris@7:     // Warpings
Chris@7:     double *x_warping;
Chris@7:     // Hanning window
Chris@7:     double *mp_HanningWindow;
Chris@7:     // FChT plan & transformed data structs
Chris@7:     double *m_absFanChirpTransform;
Chris@14:     double *m_auxFanChirpTransform;
Chris@14:     FFTReal *fft_xwarping;
Chris@7:     // GLogS
Chris@7:     double *m_glogs_f0;
Chris@7:     double *m_glogs;
Chris@10:     int *m_glogs_n;
Chris@10:     int *m_glogs_index;
Chris@10:     int *m_glogs_posint;
Chris@7:     double *m_glogs_posfrac;
Chris@7:     double *m_glogs_interp;
Chris@10:     int m_glogs_harmonic_count;
Chris@10:     int m_glogs_num_f0s;
Chris@10:     int m_glogs_init_f0s;
Chris@10:     int *m_glogs_third_harmonic_posint;
Chris@7:     double *m_glogs_third_harmonic_posfrac;
Chris@7:     double *m_glogs_third_harmonic;
Chris@10:     int *m_glogs_fifth_harmonic_posint;
Chris@7:     double *m_glogs_fifth_harmonic_posfrac;
Chris@7:     double *m_glogs_fifth_harmonic;
Chris@7:     double *m_glogs_f0_preference_weights;
Chris@7:     double *m_glogs_median_correction;
Chris@7:     double *m_glogs_sigma_correction;
Chris@0:     // auxiliar methods
Chris@7:     void design_GLogS();
Chris@0:     void design_FChT();
Chris@0:     void define_warps_linear_chirps(double *, double *);
Chris@0:     void design_warps(double *, double *, double *);
Chris@0:     void design_LPF();
Chris@0:     void clean_LPF();
Chris@0:     void apply_LPF();
Chris@7:     void design_time_window();
Chris@0: };
Chris@0: 
Chris@0: 
Chris@0: #endif