def pyin(filename,
inputSampleRate=44100,
channels=1,
frame_length=2048,
hop_length=256,
lowAmp=0.25,
onsetSensitivity=0.7,
pruneThresh=0.1):
audio, fs = librosa.load(filename, sr=inputSampleRate)
pYinInst = pYINmain.PyinMain()
pYinInst.initialise(channels=channels,
inputSampleRate=inputSampleRate,
stepSize=hop_length,
blockSize=frame_length,
lowAmp=lowAmp,
onsetSensitivity=onsetSensitivity,
pruneThresh=pruneThresh)
pYinInst.m_yin.m_yinBufferSize = int(frame_length / 2)
all_frames = librosa.util.frame(audio,
frame_length=frame_length,
hop_length=hop_length)
rmslist = []
for frame in all_frames.T:
fs, rms = pYinInst.process(frame)
rmslist.append(rms)
monoPitch = pYinInst.getSmoothedPitchTrack()
pitchlist = [ii.values for ii in fs.m_oSmoothedPitchTrack]
return np.array(pitchlist), np.array(rmslist)
def pYINPtNote(filename1,fs=44100,frameSize=2048,hopSize=256):
'''
Given filename, return pitchtrack and note transcription track
:param filename1:
:param fs:
:param frameSize:
:param hopSize:
:return:
'''
# initialise
pYinInst = pYINmain.PyinMain()
pYinInst.initialise(channels = 1, inputSampleRate = fs, stepSize = hopSize, blockSize = frameSize,
lowAmp = 0.25, onsetSensitivity = 0.7, pruneThresh = 0.1)
# frame-wise calculation
audio = ess.MonoLoader(filename = filename1, sampleRate = fs)()
# rms mean
# rms = []
# for frame in ess.FrameGenerator(audio, frameSize=frameSize, hopSize=hopSize):
# rms.append(RMS(frame, frameSize))
# rmsMean = np.mean(rms)
# print 'rmsMean', rmsMean
for frame in ess.FrameGenerator(audio, frameSize=frameSize, hopSize=hopSize):
fs = pYinInst.process(frame)
# calculate smoothed pitch and mono note
monoPitch = pYinInst.getSmoothedPitchTrack()
# output smoothed pitch track
print('pitch track')
for ii in fs.m_oSmoothedPitchTrack:
print(ii.values)
print('\n')
fs = pYinInst.getRemainingFeatures(monoPitch)
# output of mono notes,
# column 0: frame number,
# column 1: pitch in midi numuber, this is the decoded pitch
# column 2: attack 1, stable 2, silence 3
print('mono note decoded pitch')
for ii in fs.m_oMonoNoteOut:
print(ii.frameNumber, ii.pitch, ii.noteState)
print('\n')
print('note pitch tracks')
for ii in fs.m_oNotePitchTracks:
print(ii)
print('\n')
# median pitch in Hz of the notes
print('median note pitch')
for ii in fs.m_oNotes:
print(ii.values)
print('\n')
def plot(self,canvas,a,b):
global filename
filename = tkFileDialog.askopenfilename()
f_s, afAudioData = wavfile.read(filename)
afAudioData = afAudioData.astype(np.float)
#convert to float
# Normalizing and Downsizing File
if (len(afAudioData.shape) > 1):
afAudioData = afAudioData.mean(axis = 1)
if (len(afAudioData) > 1):
afAudioData = afAudioData/max(abs(afAudioData))
c = ['r','b','g'] # plot marker colors
a.clear() # clear axes from previous plot
a.plot((np.arange(0,len(afAudioData)).astype(float))/f_s, afAudioData)
# create pYIN Object
pYinInst = pYINmain.PyinMain()
pYinInst.initialise(channels = 1, inputSampleRate = f_s, stepSize = hopSize, blockSize = frameSize, lowAmp = 0.25, onsetSensitivity = 0.7, pruneThresh = 0.1)
fstart_vec = np.array(range(0,len(afAudioData)-frameSize, hopSize))
for fstart in fstart_vec:
if fstart+frameSize <= len(afAudioData):
frame = afAudioData[fstart:fstart+frameSize]
else:
frame = afAudioData[fstart:end]
fs = pYinInst.process(frame)
monoPitch = pYinInst.getSmoothedPitchTrack()
pitches = []
for ii in fs.m_oSmoothedPitchTrack:
pitches = np.concatenate((pitches,ii.values), axis=0)
p = 69 + 12 * np.log2(abs(pitches)/440)
x = fstart_vec.astype(float)/float(f_s)
b.plot(x,p)
# b.yticks(np.arange(min(pitches), max(pitches)+1, 1.0))
#b.yaxis.set_major_locator(plt.MaxNLocator(3))
canvas.draw()
sys.path.append(srcpath)
import pYINmain
import essentia.standard as ess
import numpy as np
from YinUtil import RMS
if __name__ == "__main__":
# initialise
filename1 = srcpath + '/testAudioLong.wav'
fs = 44100
frameSize = 2048
hopSize = 256
pYinInst = pYINmain.PyinMain()
pYinInst.initialise(channels=1,
inputSampleRate=fs,
stepSize=hopSize,
blockSize=frameSize,
lowAmp=0.25,
onsetSensitivity=0.7,
pruneThresh=0.1)
# frame-wise calculation
audio = ess.MonoLoader(filename=filename1, sampleRate=fs)()
# rms mean
# rms = []
# for frame in ess.FrameGenerator(audio, frameSize=frameSize, hopSize=hopSize):
# rms.append(RMS(frame, frameSize))