diff options
| -rw-r--r-- | simpl/base.pyx | 30 | ||||
| -rw-r--r-- | src/simpl/base.cpp | 78 | ||||
| -rw-r--r-- | src/simpl/base.h | 13 | ||||
| -rw-r--r-- | tests/test_base.py | 25 |
4 files changed, 98 insertions, 48 deletions
diff --git a/simpl/base.pyx b/simpl/base.pyx index 2fab453..3f811a1 100644 --- a/simpl/base.pyx +++ b/simpl/base.pyx @@ -32,7 +32,7 @@ cdef extern from "../src/simpl/base.h" namespace "simpl": void max_peaks(int new_max_peaks) void add_peak(c_Peak* peak) c_Peak* peak(int peak_number) - void clear_peaks() + void clear() # partials # int num_partials() @@ -74,9 +74,8 @@ cdef extern from "../src/simpl/base.h" namespace "simpl": void min_peak_separation(double new_min_peak_separation) int num_frames() c_Frame* frame(int frame_number) - vector[c_Frame*] frames() - # Peaks* find_peaks_in_frame(const Frame& frame) - # Frames* find_peaks(number* audio) + vector[c_Peak*] find_peaks_in_frame(c_Frame* frame) + vector[c_Frame*] find_peaks(int audio_size, double* audio) cdef class Peak: @@ -122,7 +121,7 @@ cdef class Frame: self.thisptr = new c_Frame() self.created = True else: - self.create = False + self.created = False def __dealloc__(self): if self.created: @@ -158,8 +157,8 @@ cdef class Frame: def __set__(self, peaks): self.add_peaks(peaks) - def clear_peaks(self): - self.thisptr.clear_peaks() + def clear(self): + self.thisptr.clear() # audio buffers property size: @@ -242,11 +241,26 @@ cdef class PeakDetection: def frame(self, int i): cdef c_Frame* c_f = self.thisptr.frame(i) - f = Frame(False) + f = Frame(None, False) f.set_frame(c_f) + return f property frames: def __get__(self): return [self.frame(i) for i in range(self.thisptr.num_frames())] def __set__(self, f): raise Exception("NotImplemented") + + def find_peaks_in_frame(self, Frame frame not None): + peaks = [] + cdef vector[c_Peak*] c_peaks = self.thisptr.find_peaks_in_frame(frame.thisptr) + for i in range(c_peaks.size()): + peak = Peak(False) + peak.set_peak(c_peaks[i]) + peaks.append(peak) + return peaks + + def find_peaks(self, np.ndarray[dtype_t, ndim=1] audio): + frames = [] + cdef vector[c_Frame*] c_frames = self.thisptr.find_peaks(len(audio), <double*> audio.data) + return frames diff --git a/src/simpl/base.cpp b/src/simpl/base.cpp index afffcbb..3f6c093 100644 --- a/src/simpl/base.cpp +++ b/src/simpl/base.cpp @@ -1,9 +1,9 @@ -#include <iostream> #include "base.h" using namespace std; using namespace simpl; + // --------------------------------------------------------------------------- // Peak // --------------------------------------------------------------------------- @@ -55,6 +55,7 @@ bool Peak::is_free(const string direction) return true; } + // --------------------------------------------------------------------------- // Frame // --------------------------------------------------------------------------- @@ -103,7 +104,7 @@ void Frame::max_peaks(int new_max_peaks) { _max_peaks = new_max_peaks; - // potentially losing data here but the user shouldn't really do this + // TODO: potentially losing data here, should prevent or complain if((int)_peaks.size() > _max_peaks) { _peaks.resize(_max_peaks); @@ -128,7 +129,7 @@ Peak* Frame::peak(int peak_number) return _peaks[peak_number]; } -void Frame::clear_peaks() +void Frame::clear() { _peaks.clear(); } @@ -231,6 +232,7 @@ number* Frame::synth_residual() return _synth_residual; } + // --------------------------------------------------------------------------- // PeakDetection // --------------------------------------------------------------------------- @@ -249,6 +251,16 @@ PeakDetection::PeakDetection() PeakDetection::~PeakDetection() { + clear(); +} + +void PeakDetection::clear() +{ + for(int i = 0; i < _frames.size(); i++) + { + delete _frames[i]; + } + _frames.clear(); } @@ -346,15 +358,15 @@ Frame* PeakDetection::frame(int frame_number) return _frames[frame_number]; } -Frames* PeakDetection::frames() +Frames PeakDetection::frames() { - return &_frames; + return _frames; } // Find and return all spectral peaks in a given frame of audio -Peaks* PeakDetection::find_peaks_in_frame(const Frame& frame) +Peaks PeakDetection::find_peaks_in_frame(Frame* frame) { - Peaks* peaks = new Peaks(); + Peaks peaks; return peaks; } @@ -362,30 +374,30 @@ Peaks* PeakDetection::find_peaks_in_frame(const Frame& frame) // If the signal contains more than 1 frame worth of audio, it will be broken // up into separate frames, each containing a std::vector of peaks. // Frames* PeakDetection::find_peaks(const samples& audio) -Frames* PeakDetection::find_peaks(number* audio) -{ - // _frames.clear(); - // unsigned int pos = 0; - // while(pos < audio.size()) - // { - // // get the next frame size - // if(!_static_frame_size) - // { - // _frame_size = next_frame_size(); - // } - - // // get the next frame - // Frame f = Frame(_frame_size); - // f.audio(audio, pos); - - // // find peaks - // Peaks* peaks = find_peaks_in_frame(f); - // f.add_peaks(peaks); - // delete peaks; - - // _frames.push_back(f); - // pos += _hop_size; - // } - - // return &_frames; +Frames PeakDetection::find_peaks(int audio_size, number* audio) +{ + clear(); + unsigned int pos = 0; + + while(pos < audio_size - _hop_size) + { + // get the next frame size + if(!_static_frame_size) + { + _frame_size = next_frame_size(); + } + + // get the next frame + Frame* f = new Frame(_frame_size); + f->audio(&audio[pos]); + + // find peaks + Peaks peaks = find_peaks_in_frame(f); + f->add_peaks(&peaks); + + _frames.push_back(f); + pos += _hop_size; + } + + return _frames; } diff --git a/src/simpl/base.h b/src/simpl/base.h index 9f12482..d460d0c 100644 --- a/src/simpl/base.h +++ b/src/simpl/base.h @@ -12,6 +12,7 @@ namespace simpl typedef double number; + // --------------------------------------------------------------------------- // Peak // @@ -41,6 +42,7 @@ public: typedef std::vector<Peak*> Peaks; + // --------------------------------------------------------------------------- // Partial // --------------------------------------------------------------------------- @@ -48,6 +50,7 @@ class Partial {}; typedef std::vector<Partial*> Partials; + // --------------------------------------------------------------------------- // Frame // @@ -85,7 +88,7 @@ public: void add_peak(Peak* peak); void add_peaks(Peaks* peaks); Peak* peak(int peak_number); - void clear_peaks(); + void clear(); Peaks::iterator peaks_begin(); Peaks::iterator peaks_end(); @@ -111,6 +114,7 @@ public: typedef std::vector<Frame*> Frames; + // --------------------------------------------------------------------------- // PeakDetection // @@ -133,6 +137,7 @@ private: public: PeakDetection(); virtual ~PeakDetection(); + void clear(); int sampling_rate(); void sampling_rate(int new_sampling_rate); @@ -153,15 +158,15 @@ public: void min_peak_separation(number new_min_peak_separation); int num_frames(); Frame* frame(int frame_number); - Frames* frames(); + Frames frames(); // Find and return all spectral peaks in a given frame of audio - virtual Peaks* find_peaks_in_frame(const Frame& frame); + virtual Peaks find_peaks_in_frame(Frame* frame); // Find and return all spectral peaks in a given audio signal. // If the signal contains more than 1 frame worth of audio, it will be broken // up into separate frames, with an array of peaks returned for each frame. - virtual Frames* find_peaks(number* audio); + virtual Frames find_peaks(int audio_size, number* audio); }; } // end of namespace Simpl diff --git a/tests/test_base.py b/tests/test_base.py index 8b00d33..aade0df 100644 --- a/tests/test_base.py +++ b/tests/test_base.py @@ -1,5 +1,8 @@ -import simpl.base as base +import os import numpy as np +import scipy.io.wavfile as wavfile +from nose.tools import assert_almost_equals +import simpl.base as base class TestFrame(object): @@ -39,11 +42,27 @@ class TestFrame(object): assert f.peak(0).amplitude == p.amplitude assert f.peaks[0].amplitude == p.amplitude - f.clear_peaks() + f.clear() assert f.num_peaks == 0 class TestPeakDetection(object): + float_precision = 5 + frame_size = 512 + hop_size = 512 + audio_path = os.path.join( + os.path.dirname(__file__), 'audio/flute.wav' + ) + + @classmethod + def setup_class(cls): + cls.audio = wavfile.read(cls.audio_path)[1] + cls.audio = np.asarray(cls.audio, dtype=np.double) + cls.audio /= np.max(cls.audio) + def test_peak_detection(self): pd = base.PeakDetection() - print pd.frames + pd.find_peaks(self.audio) + + assert len(pd.frames) == len(self.audio) / self.hop_size + assert len(pd.frames[0].peaks) == 0 |