PeriodicityPitch: Periodicity Pitch Image

In m-vidal/eaR: Perception-Based Auditory and Music Analysis

Periodicity Pitch Image

Description

This function calculates the periodicity pitch image (PPI) from the ANI matrix obtained with the function CalcANI. The periodicity pitch is defined as the summed autocorrelation over bandpass filtered fluctuations on the auditory channels. Since the output of the auditory model gives the envelopes of the neural firing probabilities (< 1250Hz), it suffices to apply a low-pass filter to obtain the pitch in a reliable frequency range (80-1250 Hz).

Let F_{j} denote a second-order Butterworth filter with a cutoff frequency of 80 Hz. Then, the filtered channels are \tilde{A}_{j}=F_{j}[A_j], where A_j\ (j=1,…,m) denote the channels of the ANI matrix. The lower limit of 80 Hz accounts for the fact that for smaller frequencies, the pitch becomes more a sensation of textural properties. The higher limit of 1250 Hz is related to the limits of neural synchronization. Beyond about 1250 Hz, the neurons are no longer able to follow the exact period of the signal very accurately, and periodicity pitch becomes unreliable. Only the lowest frequency can be changed by inLowFrequency. Secondly, a frame-based convolution analysis and a coincidence mechanism is conducted by the summation of the convolution results over all channels.

Usage

PeriodicityPitch(inANIObj, inLowFrequency = 80,
                 inFrameWidth = 0.064, inFrameStepSize = 0.010)

Arguments

inANIObj	an object of class "AI". It must contain the output of CalcANI function: a matrix of size n x m where n is the number of auditory channels (<= 40) and m is the number of samples. the sample frequency of the input signal (in Hz).
inLowFrequency	cutoff frequency (in Hz) of a first order lowpass filter applied before calculating the autocorrelation. If empty or not specified, 80 is used by default.
inFrameWidth	width of the frame used for the accumulation of the autocorrelation (in s). If empty or not specified, 0.064 is used by default.
inFrameStepSize	stepsize or time interval between two inFrameWidth (in s). If empty or not specified, 0.010 is used by default.

Details

As for any frame-based function, the first value of the output signal is the value calculated for the first complete frame in the input signal. Thus, if you have an input signal of length 1 s at a sampling frequency of 1000 Hz, a frame width of 0.050 s and a frame step size of 0.010 s, then there will be ceiling(((1 - 0.050)*1000 + 1)/(0.010*1000)) = 96 values in the output signal. The first value corresponds to the first complete frame (the interval 0 to 0.050 s).

Value

An object of class "AI", which is a list with the following elements:

PeriodicityPitchImage	periodicity pitch: a matrix of size inFrameWidth * length(inMatrix) / outSampleFreq
SampleFreq	sampling rate, equal to inSampleFreq/inFrameStepSize (in Hz).
Periods	analyzed periods (in s).
BPANI	bandpass filtered auditory nerve images (at the original sample frequency).

Author(s)

Marc Vidal (R version). Based on the original code from IPEM Toolbox.

Examples

## Not run: probe <- ShepardTone(293.66, 1, indBLevel = -20)
s <- c(SchumannKurioseGeschichte, numeric(2205), probe)
ANIs <- CalcANI(s, 22050)
PPs <- PeriodicityPitch(ANIs)
## End(Not run)

m-vidal/eaR documentation built on Nov. 18, 2022, 3:55 p.m.