pyACA.PitchTimeAmdf Namespace Reference

Functions
	PitchTimeAmdf (x, iBlockLength, iHopLength, f_s)
	computes f0 via the lag of the amdf function

Function Documentation

PitchTimeAmdf()

PitchTimeAmdf	(		x,
			iBlockLength,
			iHopLength,
			f_s )

computes f0 via the lag of the amdf function

Parameters

x	array with floating point audio data (dimension samples x channels)
iBlockLength	internal block length
iHopLength	internal hop length
f_s	sample rate of audio data

Returns

f_0: fundamental frequency (in Hz)

t: time stamp

Definition at line 18 of file PitchTimeAmdf.py.

def PitchTimeAmdf(x, iBlockLength, iHopLength, f_s):
    def computeAmdf(x, eta_max):
        K = x.shape[0]
 
        if K <= 0:
            return 0
 
        afAmdf = np.ones(K)
 
        for eta in range(0, np.min([K, eta_max + 1])):
            afAmdf[eta] = np.sum(np.abs(x[np.arange(0, K - 1 - eta)] - x[np.arange(eta + 1, K)])) / K
 
        return afAmdf
 
 
    # initialize
    f_max = 2000
    f_min = 50
 
    # block audio data
    x_b, t = ToolBlockAudio(x, iBlockLength, iHopLength, f_s)
    iNumOfBlocks = x_b.shape[0]
 
    # allocate memory
    f_0 = np.zeros(iNumOfBlocks)
 
    eta_min = np.floor(f_s / f_max).astype(int)
    eta_max = np.floor(f_s / f_min).astype(int)
 
    for n, block in enumerate(x_b):
 
        # calculate the amdf if non zero
        if not block.sum():
            continue
        else:
            afCorr = computeAmdf(block, eta_max)
 
        # find the coefficients specified in eta
        f_0[n] = np.argmin(afCorr[np.arange(eta_min + 1, afCorr.size)]) + 1
 
        # convert to Hz
        f_0[n] = f_s / (f_0[n] + eta_min + 1)
 
    return f_0, t