def extract(self, segmentation, feature, **kwargs):
"""
Given a segmentation and a feature class (plus **kwargs), extract feature for each segment in the segmentation.
"""
if type(segmentation.media) is not str:
raise TypeError("Only file segmentation extraction is currently supported.")
x,SR,fmt = sound.wavread(segmentation.media, last=1)
for seg in segmentation:
x, xsr, fmt = sound.wavread(segmentation.media,
first=int(seg.time_span.start_time*SR), last=int(seg.time_span.duration*SR))
f = feature(x, **kwargs)
seg.features.append(f)
def extract(self, segmentation, feature, **kwargs):
"""
Given a segmentation and a feature class (plus **kwargs), extract feature for each segment in the segmentation.
"""
if type(segmentation.media) is not str:
raise TypeError(
"Only file segmentation extraction is currently supported.")
x, SR, fmt = sound.wavread(segmentation.media, last=1)
for seg in segmentation:
x, xsr, fmt = sound.wavread(segmentation.media,
first=int(seg.time_span.start_time * SR), last=int(seg.time_span.duration * SR))
f = feature(x, **kwargs)
seg.features.append(f)
def play_segs(self, k):
"""
Play audio segments corresponding to cluster k (zero-based index)
"""
if k is None:
raise TypeError("play_segs requires an integrer cluster index: k")
clusters = self.assigns[self.diffs[:-1]]
sr = self.feature.sample_rate
for seg in np.where(clusters==k)[0]:
print self.segmentation[seg]
x = sound.wavread(self.media, first=int(self.segmentation[seg].time_span.start_time*sr), last=int(self.segmentation[seg].time_span.duration*sr))
sound.play(x[0].T, sr)
def play_segs(self, k):
"""
Play audio segments corresponding to cluster k (zero-based index)
"""
if k is None:
raise TypeError("play_segs requires an integrer cluster index: k")
clusters = self.assigns[self.diffs[:-1]]
sr = self.feature.sample_rate
for seg in np.where(clusters == k)[0]:
print(self.segmentation[seg])
x = sound.wavread(self.media, first=int(self.segmentation[
seg].time_span.start_time * sr), last=int(self.segmentation[seg].time_span.duration * sr))
sound.play(x[0].T, sr)
import numpy as np
import scipy.signal as signal
import sound
import matplotlib.pyplot as plt
# loading sound file as array
[s, r] = sound.wavread('Track32.wav')
# Taking first chanel of audio
data = s[:, 0]
# filter
#First is a Low Pass Filter
h1 = signal.remez(64, [0, 0.0625, 0.0627, 0.5], [1, 0],
[1, 100]) # nyquist frequency normalized to 0.5
#Band Pass filters
h2 = signal.remez(64, [0, 0.0625, 0.0627, 0.125, 0.127, 0.5], [0, 1, 0],
[100, 1, 100])
h3 = signal.remez(64, [0, 0.125, 0.127, 0.1875, 0.1877, 0.5], [0, 1, 0],
[100, 1, 100])
h4 = signal.remez(64, [0, 0.1875, 0.1877, 0.25, 0.27, 0.5], [0, 1, 0],
[100, 1, 100])
h5 = signal.remez(64, [0, 0.25, 0.27, 0.3125, 0.3127, 0.5], [0, 1, 0],
[100, 1, 100])
h6 = signal.remez(64, [0, 0.3125, 0.3127, 0.375, 0.377, 0.5], [0, 1, 0],
[100, 1, 100])
h7 = signal.remez(64, [0, 0.375, 0.377, 0.4375, 0.4377, 0.5], [0, 1, 0],
[100, 1, 100])
#High Pass filter
h8 = signal.remez(64, [0, 0.4375, 0.4377, 0.5], [0, 1], [100, 1])
#Script for calling the online MDCT implementation analysis and synthesis, for time measurement
#and as simple implementation example.
#Gerald Schuller, November 2017
from pyrecplayfastMDCT import *
import sound
import os
import time
N=1024 #Number of subbands and block size
#initialize filter bank memory:
initFB(N)
fb=np.sin(np.pi/(2*N)*(np.arange(0,int(1.5*N))+0.5))
X, fs= sound.wavread('test.wav')
print("X.shape:",X.shape)
print("fs=", fs)
X=X*1.0/2**15
blocks=len(X)/N
y=np.zeros((blocks,N))
xrek=np.zeros(blocks*N)
startime=time.time()
#analysis filter bank:
for m in range(blocks):
#print("X[m*N+np.arange(N).shape:", X[m*N+np.arange(N)].shape)
#Analysis Filter Bank:
y[m,:]=LDFB(X[m*N+np.arange(N)],fb)
#Synthesis Filter Bank:
xrek[m*N+np.arange(N)]=LDFBinv(y[m,:],fb);
# ## Check the frquency response:
# In[3]:
w, H = signal.freqz(b)
plt.plot(w, 20 * np.log10(abs(H) + 1e-6))
plt.title('Filter Magnitude Frequency Response')
plt.xlabel('Normalized Frequency')
plt.ylabel('dB')
# ## Now the **analysis filtering and down sampling:**
# In[4]:
import sound as snd
[s, rate] = snd.wavread('sndfile.wav')
print("length of sound in samples: ", len(s))
plt.plot(s)
plt.title('Original Signal')
# ## Filter implementation:
# In[5]:
filtered = signal.lfilter(b, 1, s)
print("length of filtered sound in samples: ", len(filtered))
plt.plot(filtered)
# ## Play the filtered sound:
# In[6]:
#Gerald Schuller, November 2017
from pyrecplayfastLDFB import *
import sound
import os
import time
N = 512 #Number of subbands and block size
#initialize filter bank memory:
initFB(N)
fb = np.loadtxt('fb2048t1023d512bbitcs.mat')
#N=1024:
#fb=np.loadtxt('fb4096t2047d1024bbitc.mat')
X, fs = sound.wavread('teststereo.wav')
print("X.shape:", X.shape)
print("fs=", fs)
X = X * 1.0 / 2**15
blocks = len(X) / N
y = np.zeros((blocks, 2, N))
xrek = np.zeros((blocks * N, 2))
startime = time.time()
#analysis filter bank:
for m in range(blocks):
#print("X[m*N+np.arange(N).shape:", X[m*N+np.arange(N)].shape)
#Analysis Filter Bank:
#Left channel:
y[m, 0, :] = LDFB(X[m * N + np.arange(N), 0], fb)
#Right channel:
y[m, 1, :] = LDFB(X[m * N + np.arange(N), 1], fb)
import scipy.signal
import numpy as np
import matplotlib.pyplot as plt
import sys
import cv2
CHUNK = 1024
print("filename=", sys.argv[1])
sndfile=sys.argv[1]
#sndfile="amfile.wav";
#sndfile="amfilegs.wav";
#sndfile="amrec.wav";
#read in sound file:
[AM, FS]=sound.wavread(sndfile)
#length of the sound:
lenAM=scipy.size(AM)
#compute the low pass filter coefficients,
#with 10 Hz cutoff frequency:
[b,a]=scipy.signal.iirfilter(2, 20.0/(FS/2),rp=60,btype='lowpass')
[w,H]=scipy.signal.freqz(b,a)
Ha=scipy.absolute(H)
fig=plt.figure()
#plt.plot(w,Ha)
#In dB on normalized frequency axis w:
plt.plot(w,20*np.log10(Ha))
plt.title('The frequence response of the low pass filter')
plt.xlabel('Normalized frequency (pi=Nyquist freq.)')
plt.ylabel('Magnitude response (dB)')