pyACA.FeatureSpectralSpread Namespace Reference

Functions
	FeatureSpectralSpread (X, f_s)
	computes the spectral spread from the magnitude spectrum

Function Documentation

FeatureSpectralSpread()

FeatureSpectralSpread	(		X,
			f_s )

computes the spectral spread from the magnitude spectrum

Parameters

X	spectrogram (dimension FFTLength X Observations)
f_s	sample rate of audio data

Returns

vss: spectral spread

Definition at line 13 of file FeatureSpectralSpread.py.

def FeatureSpectralSpread(X, f_s):
 
    isSpectrum = X.ndim == 1
    if isSpectrum:
        X = np.expand_dims(X, axis=1)
 
    # get spectral centroid as index
    vsc = FeatureSpectralCentroid(X, f_s) * 2 / f_s * (X.shape[0] - 1)
 
    # X = X**2 removed for consistency with book
 
    norm = X.sum(axis=0)
    norm[norm == 0] = 1
 
    # compute spread
    vss = np.zeros(X.shape[1])
    indices = np.arange(0, X.shape[0])
    for n in range(0, X.shape[1]):
        vss[n] = np.dot((indices - vsc[n])**2, X[:, n]) / norm[n]
 
    vss = np.sqrt(vss)
 
    # convert from index to Hz
    vss = vss / (X.shape[0] - 1) * f_s / 2
 
    return np.squeeze(vss) if isSpectrum else vss