pyACA.FeatureTimeRms Namespace Reference

computes the RMS of a time domain signal More...

Functions
	FeatureTimeRms (x, iBlockLength, iHopLength, f_s)
	computes the RMS of a time domain signal
	filterSP (x, alpha)

Detailed Description

computes the RMS of a time domain signal

Args: x: audio signal iBlockLength: block length in samples iHopLength: hop length in samples f_s: sample rate of audio data (unused)

Returns: vrms rms value (row 1: block-based rms, row 2: single pole approx) t time stamp

Function Documentation

FeatureTimeRms()

FeatureTimeRms	(		x,
			iBlockLength,
			iHopLength,
			f_s )

computes the RMS of a time domain signal

Parameters

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

Returns

vrms:rms value (row 1: block-based rms, row 2: single pole approx)

t: time stamp

Definition at line 30 of file FeatureTimeRms.py.

def FeatureTimeRms(x, iBlockLength, iHopLength, f_s):
 
    T_i = .3 
    alpha = 1 - np.exp(-2.2/f_s/T_i)
 
    # create blocks
    x_b, t = pyACA.ToolBlockAudio(x, iBlockLength, iHopLength, f_s)
 
    # number of results
    iNumOfBlocks = x_b.shape[0]
 
    # single pole implementation
    v_sp = filterSP(x**2, alpha)
 
    # allocate memory
    vrms = np.zeros([2, iNumOfBlocks])
 
    for n, block in enumerate(x_b):
        i_start = n * iHopLength
        i_stop = np.min([len(x), i_start + iBlockLength])
 
        # calculate the rms
        vrms[0, n] = np.sqrt(np.dot(block, block) / block.size)
        vrms[1, n] = np.max(np.sqrt(v_sp[i_start:i_stop]))
 
    # convert to dB
    epsilon = 1e-5  # -100dB
 
    vrms[vrms < epsilon] = epsilon
    vrms = 20 * np.log10(vrms)
 
    return vrms, t
 
 

Here is the call graph for this function:

filterSP()

filterSP	(		x,
			alpha )

Definition at line 64 of file FeatureTimeRms.py.

def filterSP(x, alpha):
    
    xf = np.zeros(x.shape)
    xf[0] = alpha * x[0]
 
    for i in range(1, len(x)):
        xf[i] = alpha * x[i] + (1 - alpha) * xf[i-1]
    
    return xf

Here is the caller graph for this function: