Python signal.butter函数代码示例

 def makeFiltered(self):
     filtered = np.zeros(len(self.data))
     b1, a1 = butter(1, 0.0003, 'lowpass')
     filtered = filtfilt(b1, a1, self.data)
     filtered = self.data - filtered
     b2, a2 = butter(1, 0.025, 'lowpass')
     self.filt_data = filtfilt(b2, a2, filtered)

def filter_feepmt(feep):
    """
    input: an instance of class FeePmt
    output: buttersworth parameters of the equivalent FEE filter
    """

    # high pass butterswoth filter ~1/RC
    b1, a1 = signal.butter(1, feep.freq_LHPFd, 'high', analog=False)
    b2, a2 = signal.butter(1, feep.freq_LHPFd, 'low', analog=False)

    b0 = b2*(feep.Zin/(1+feep.C1/feep.C2))+b1*(feep.Zin*feep.R1/(feep.Zin+feep.R1))
    a0 = a1

    # LPF order 1
    b1l, a1l = signal.butter(1, feep.freq_LPF1d, 'low', analog=False)
    # LPF order 4
    b2l, a2l = signal.butter(4, feep.freq_LPF2d, 'low', analog=False)
    # convolve HPF, LPF1
    a_aux = np.convolve(a0, a1l, mode='full')
    b_aux = np.convolve(b0, b1l, mode='full')
    # convolve HPF+LPF1, LPF2
    a = np.convolve(a_aux, a2l, mode='full')
    b_aux2 = np.convolve(b_aux, b2l, mode='full')
    b = feep.A2*b_aux2

    return b, a

	def __init__(self,window_size, low, high):
		fs_Hz = 250;
		fn = fs_Hz/2
		self.filtered_data = np.array((window_size,1))
		

		#######################################
		# Filter Creation
		# -------------------------------------
		#
		# Create a filter using the scipy module,
		# based on specifications suggested by
		# Pan-Tompkins (bandpass from 5-15Hz)
		#
		#
		# 1) Establish constants:
		#		a) filter_order = 2
		#		b) high pass cutoff = 15Hz
		#		c) low pass cutoff = 5Hz
		# 2) Calculate the coefficients, store in variables

		filter_order = 2
		f_high = high
		f_low = low
		self.high_pass_coefficients = signal.butter(filter_order,f_low/fn, 'high')
		self.low_pass_coefficients = signal.butter(filter_order,f_high/fn, 'low')

    def __init__(self,source, nchannels, order, fc, btype='low'):
        Wn = np.asarray(np.atleast_1d(fc)).copy() #Scalar inputs are converted to 1-dimensional arrays
        self.samplerate = source.samplerate
        Wn= Wn/float(self.samplerate)*2    # wn=1 corresponding to half the sample rate

        if btype=='low' or btype=='high':
            self.filt_b=np.zeros((nchannels,order+1))
            self.filt_a=np.zeros((nchannels,order+1))
            if len(Wn)==1:     #if there is only one Wn value for all channel just repeat it
                self.filt_b, self.filt_a = signal.butter(order, Wn, btype=btype)
                self.filt_b=np.kron(np.ones((nchannels,1)),self.filt_b)
                self.filt_a=np.kron(np.ones((nchannels,1)),self.filt_a)
            else:               #else make nchannels different filters
                for i in range((nchannels)):
                    self.filt_b[i,:], self.filt_a[i,:] = signal.butter(order, Wn[i], btype=btype)
        else:
            self.filt_b=np.zeros((nchannels,2*order+1))
            self.filt_a=np.zeros((nchannels,2*order+1))
            if Wn.ndim==1:     #if there is only one Wn pair of values for all channel just repeat it
                self.filt_b, self.filt_a = signal.butter(order, Wn, btype=btype)
                self.filt_b=np.kron(np.ones((nchannels,1)),self.filt_b)
                self.filt_a=np.kron(np.ones((nchannels,1)),self.filt_a)
            else:   
                for i in range((nchannels)):
                    self.filt_b[i,:], self.filt_a[i,:] = signal.butter(order, Wn[:,i], btype=btype)   
                    
                    
        self.filt_a=self.filt_a.reshape(self.filt_a.shape[0],self.filt_a.shape[1],1)
        self.filt_b=self.filt_b.reshape(self.filt_b.shape[0],self.filt_b.shape[1],1)  
        self.nchannels = nchannels    
        LinearFilterbank.__init__(self,source, self.filt_b, self.filt_a) 

    def __init__(self, source, nchannels, order, fc, btype="low"):
        Wn = fc.copy()
        Wn = atleast_1d(Wn)  # Scalar inputs are converted to 1-dimensional arrays
        self.samplerate = source.samplerate
        try:
            Wn = Wn / self.samplerate * 2 + 0.0  # wn=1 corresponding to half the sample rate
        except DimensionMismatchError:
            raise DimensionMismatchError("Wn must be in Hz")

        if btype == "low" or btype == "high":
            self.filt_b = zeros((nchannels, order + 1))
            self.filt_a = zeros((nchannels, order + 1))
            if len(Wn) == 1:  # if there is only one Wn value for all channel just repeat it
                self.filt_b, self.filt_a = signal.butter(order, Wn, btype=btype)
                self.filt_b = kron(ones((nchannels, 1)), self.filt_b)
                self.filt_a = kron(ones((nchannels, 1)), self.filt_a)
            else:  # else make nchannels different filters
                for i in xrange((nchannels)):
                    self.filt_b[i, :], self.filt_a[i, :] = signal.butter(order, Wn[i], btype=btype)
        else:
            self.filt_b = zeros((nchannels, 2 * order + 1))
            self.filt_a = zeros((nchannels, 2 * order + 1))
            if Wn.ndim == 1:  # if there is only one Wn pair of values for all channel just repeat it
                self.filt_b, self.filt_a = signal.butter(order, Wn, btype=btype)
                self.filt_b = kron(ones((nchannels, 1)), self.filt_b)
                self.filt_a = kron(ones((nchannels, 1)), self.filt_a)
            else:
                for i in xrange((nchannels)):
                    self.filt_b[i, :], self.filt_a[i, :] = signal.butter(order, Wn[:, i], btype=btype)

        self.filt_a = self.filt_a.reshape(self.filt_a.shape[0], self.filt_a.shape[1], 1)
        self.filt_b = self.filt_b.reshape(self.filt_b.shape[0], self.filt_b.shape[1], 1)
        self.nchannels = nchannels
        LinearFilterbank.__init__(self, source, self.filt_b, self.filt_a)

 def butter_high_low_pass(lowcut, highcut, sampling_rate, order=5):
     nyq_freq = sampling_rate*0.5
     lower_bound = lowcut/nyq_freq
     higher_bound = highcut/nyq_freq
     b_high, a_high = butter(order, lower_bound, btype='high')
     b_low, a_low = butter(order, higher_bound, btype='low')
     return b_high, a_high, b_low, a_low

  def filter_data(self,eeg_data):
    #FILTER CONSTANTS
    fs = self.fs
    fn = self.fn
    filter_order = 2   #2nd order filter
    f_high = 50
    f_low = 5
    wn = [59,61]       #Nyquist filter window

    [b,a] = signal.butter(filter_order,f_high/fn, 'low')
    [b1,a1] = signal.butter(filter_order,f_low/fn, 'high')
    [bn,an] = signal.butter(4,[x/fn for x in wn], 'stop')

    filtered_eeg = []
    spectogram = []
    notched = []
    high_passed = []
    low_passed = []
    print(eeg_data)
    for i in range(len(eeg_data[0])):
      channel =  eeg_data[:,i]
      high_passed = signal.filtfilt(b1,a1,channel);        # high pass filter
      low_passed = signal.filtfilt(b,a,high_passed);                # low pass filter
      y = signal.filtfilt(bn,an,low_passed);        # notch filter
      filtered_eeg.append(y);
    self.filtered_eeg = filtered_eeg
    return filtered_eeg


# if __name__ == '__main__':
#   main()

    def filter(self):
        path = os.getcwd()+'/trialGraspEventDetection_dataFiles'
        self.Fgr = np.sum(self.values[:, 9:15], axis=1)  # SAI
        self.Fgl = np.sum(self.values[:, 0:7], axis=1)  # SAI

        # can use this to plot in matlab graspeventdetection_plot.m
        np.savetxt(path+'/SAI_Fgr.txt', self.Fgr)
        # can use this to plot in matlab
        np.savetxt(path+'/SAI_Fgl.txt', self.Fgl)

        # 0.55*pi rad/samples
        b1, a1 = signal.butter(1, 0.55, 'high', analog=False)
        self.f_acc_x = signal.lfilter(b1, a1, self.acc_x, axis=-1, zi=None)
        self.f_acc_y = signal.lfilter(b1, a1, self.acc_y, axis=-1, zi=None)
        self.f_acc_z = signal.lfilter(b1, a1, self.acc_z, axis=-1, zi=None)
        # self.f_eff = signal.lfilter(b1, a1, self.eff, axis=-1, zi=None)
        # type(eff)
        self.FAII = np.sqrt(np.square(self.f_acc_x) +
                            np.square(self.f_acc_y) +
                            np.square(self.f_acc_z))
        # can use this to plot in matlab
        np.savetxt(path+'/FAII.txt', self.FAII)

        # subtract base values from the values array
        self.values1 = self.values - self.values.min(axis=0)
        # pass the filter for each sensor
        self.fvalues1 = np.zeros(self.values1.shape)
        # 0.48*pi rad/samples
        b, a = signal.butter(1, 0.48, 'high', analog=False)
        for i in range(16):
            self.fvalues1[:, i] = signal.lfilter(b, a, self.values1[:, i],
                                                 axis=-1, zi=None)
        self.FAI = np.sum(self.fvalues1, axis=1)
        # can use this to plot in matlab
        np.savetxt(path+'/FAI.txt', self.FAI)

def butter_bandpass(fs, lowcut=None, highcut=None, order=10):
    """
    Parameters
    ----------
    fs : float
        Sample Rate
    lowcut : float
        Low pass frequency cutoff
    highcut : float
        High pass frequency cutoff
    order : int
        Butterworth filter order [Default = 10, i.e., 10th order Butterworth filter]

    Reference
    ---------
    http://scipy-cookbook.readthedocs.io/items/ButterworthBandpass.html
    """
    from scipy.signal import butter
    nyq = 0.5 * fs
    if lowcut and highcut:
        high = highcut / nyq
        low = lowcut / nyq
        sos = butter(order, [low, high], analog=False, btype='band', output='sos')
    elif highcut:
        high = highcut / nyq
        sos = butter(order, high, btype='low', output='sos')
    elif lowcut:
        low = lowcut / nyq
        sos = butter(order, low, btype='high', output='sos')
    else:
        print('Error -- must supply lowcut, highcut or both')
        sos = None

    return sos

def temporally_filter(spatial, tfiltN, flash):
    print("Temporally Filtering")
    padlen=0
    if tfiltN[2]==1: #%if only high pass temporal filter
        [b,a]= butter(tfiltN[0],tfiltN[1],btype='highpass')
        padlen=3
    else: #%if band pass temporal filter
        [b,a]=butter(tfiltN[0],[tfiltN[1],tfiltN[2]], btype='bandpass')
        padlen=6
    temporal=np.array(spatial, dtype = np.float32)

    if len(flash) == 2:
        temporal[flash[0]:flash[1]+1,:,:]=np.mean(spatial[4:flash[0]+1, :, :], 0)
        bg=np.mean(temporal[:flash[0] + 1, :, :],0)
    else:
        bg = np.min(temporal, 0)

    for t in np.arange(len(temporal)):
        temporal[t, :, :] -= bg

    #temporal[0, :, :] = 0
    p = 0.0
    for y in range(np.size(temporal, 1)):
        for x in range(np.size(temporal, 2)):
            temporal[:, y, x]=filtfilt(b,a, temporal[:, y, x], padlen=padlen)
        if (100 * y / np.size(temporal, 1)) > p:
            p = (100 * y / np.size(temporal, 1))
            print("  %d%%\r" % (100 * y / np.size(temporal, 1))),
    print('Temporally Filtered')
    return temporal

    def __init__(self, channels, sample_rate, leds=None):
        self.leds = leds # Not needed if just plotting
        self.channels = channels
        self.sample_rate = sample_rate
        self.nyquist = float(sample_rate) / 2

        # Filter order - higher the order the sharper
        # the curve
        order = 3

        # Cut off frequencies:
        # Low pass filter
        cutoff = 200 / self.nyquist
        # Numerator (b) and denominator (a)
        # polynomials of the filter. 
        b, a = butter(order, cutoff, btype='lowpass')
        self.low_b = b
        self.low_a = a

        # High pass filter
        cutoff = 4000 / self.nyquist
        b, a = butter(order, cutoff, btype='highpass')
        self.high_b = b
        self.high_a = a
        
        # Keep track of max brightness for each
        # colour
        self.max = [0.0, 0.0, 0.0]
        # Make different frequencies fall faster
        # bass needs to be punchy.
        self.fall = [15.0, 2.5, 5.0]

def prepare_audio_filters():
    tf_rangel = 100000
    tf_rangeh = 170000

    # audio filters
    tf = SysParams["audio_lfreq"]
    N, Wn = sps.buttord(
        [(tf - tf_rangel) / (freq_hz / 2.0), (tf + tf_rangel) / (freq_hz / 2.0)],
        [(tf - tf_rangeh) / (freq_hz / 2.0), (tf + tf_rangeh) / (freq_hz / 2.0)],
        5,
        15,
    )
    Faudl = filtfft(sps.butter(N, Wn, btype="bandpass"))

    tf = SysParams["audio_rfreq"]
    N, Wn = sps.buttord(
        [(tf - tf_rangel) / (freq_hz / 2.0), (tf + tf_rangel) / (freq_hz / 2.0)],
        [(tf - tf_rangeh) / (freq_hz / 2.0), (tf + tf_rangeh) / (freq_hz / 2.0)],
        5,
        15,
    )
    Faudr = filtfft(sps.butter(N, Wn, btype="bandpass"))

    N, Wn = sps.buttord(0.016 / (afreq / 2.0), 0.024 / (afreq / 2.0), 5, 15)
    FiltAPost = filtfft(sps.butter(N, Wn))

    N, Wn = sps.buttord(3.1 / (freq / 2.0), 3.5 / (freq / 2.0), 1, 20)
    SysParams["fft_audiorf_lpf"] = Faudrf = filtfft(sps.butter(N, Wn, btype="lowpass"))

    SysParams["fft_audiolpf"] = FiltAPost  # * FiltAPost * FiltAPost

    SysParams["fft_audio_left"] = Faudrf * Faudl * fft_hilbert
    SysParams["fft_audio_right"] = Faudrf * Faudr * fft_hilbert

def butter_bandpass(lowcut, highcut, samplingrate, order=4):
    nyq = 0.5 * samplingrate
    low = lowcut / nyq
    high = highcut / nyq
    print high, low
    if high >=1. and low == 0.:
        b = np.array([1.])
        a = np.array([1.])

    elif high < 0.95 and low > 0. :
        wp = [1.05*low,high-0.05]
        ws = [0.95*low,high+0.05]
 
        print wp,ws
        order,wn = buttord(wp,ws,0., 30.)
        b, a = butter(order, wn, btype='band')
    
    elif high>= 0.95:
        print 'highpass',low,1.2*low,0.8*low
        order,wn = buttord( 15*low,0.05*low,gpass=0.0, gstop=10.0)
        print order,wn
        b, a = butter(order, wn, btype='high')
    elif low <= 0.05:
        print 'lowpass',high
        order,wn = buttord( high-0.05,high+0.05,gpass=0.0, gstop=10.0)
        b, a = butter(order, wn, btype='low')

    return b, a

    def __init__(self, fs, filter_type):

        self.filter_type = filter_type # Of two types. Alpha and Bandpass
        self.fs = fs # sample frequency: 220hz
        
        if self.filter_type == 'alpha':
            # butter = butterworth filter function
            # An optimal (in one way -- smoothness) for designing a filter. Simple and popular.
            # signal.butter is from scipy. Is there an equivalent in Java? Could be developed in Python and copied
            # and pasted into Java. The values of these arrays is really only dependent on sampling frequency and thus
            # likely to be the same for all uses of the app
            # Returns two arrays of floats. Could b
            self.b, self.a = signal.butter(5# order of filter. More coefficients = better filtration at cost of speed
                , np.array([8.,12.])# boundaries of filter (8-12hz for Alpha
                /(self.fs/2.) # /2 because this function needs normalized frequency
                , 'bandpass')
        elif self.filter_type == 'bandpass':
            self.b, self.a = signal.butter(5, np.array([2., 36.] # bandpass filter has different freq cutoffs
                )/(self.fs/2.), 'bandpass')
        else:
            print('Filter type ''%s'' not supported.'%self.filter_type)
            return

        self.nb = len(self.b)
        self.na = len(self.a)

def generate_noise(D,N):
  """Generate data for the changepoint detection. Data can either be of type 0
  or type 1, but when it's a combination fo both, we define a target label
  Input
  - D,N Dimenstionality arguments D dimensions over N samples
  Output
  - Data in format
  X is a matrix in R^{N x D}
  y is a matrix in R^{N,} not to donfuse with {N,1}"""
  #Check if we have even D, so we can split the array in future
  assert D%2 == 0, 'We need even number of dimensions'
  ratioP = 0.5   #balance of targets
  X = np.random.randn(N,D)
  y = np.zeros(N)
  mark = np.zeros(N)
  #Generate two filter cofficients
  filters = {}
  filters['b1'],filters['a1'] = signal.butter(4,2.0*cutoff1/fs,btype='lowpass')
  filters['b0'],filters['a0'] = signal.butter(4,2.0*cutoff0/fs,btype='lowpass')
  for i in xrange(N):
    if np.random.rand() > 0.5:	#Half of the samples will have changepoint, other half wont
      Dcut = np.random.randint(pattern_len,D-pattern_len)
      signalA = signal.filtfilt(filters['b1'],filters['a1'],X[i])
      signalB = signal.filtfilt(filters['b0'],filters['a0'],X[i])
      X[i] = np.concatenate((signalA[:Dcut],signalB[Dcut:]),axis=0)    #Concatenate the two signals
      if True:  #Boolean: do you want to introduce a pattern at the changepoint?
        Dstart = int(Dcut - pattern_len/2)
        X[i,Dstart:Dstart+pattern_len] = pattern
      y[i] = 1		#The target label
      mark[i] = Dcut
    else:
      mode = int(np.random.rand()>ratioP)
      X[i] = signal.filtfilt(filters['b'+str(mode)],filters['a'+str(mode)],X[i])
      y[i] = 0		#The target label
  return X,y,mark