A Novel Normalization Method for Autocorrelation Function for Pitch Detection and for Speech Activity Detection

Abstract

Autocorrelation functions (ACF) have been used in various pitch detection algorithms (PDA) and voicing-feature based speech activity detection (SAD) techniques. Speech is assumed to be stationary over a short-term window, and a Hanning window is typically applied in the calculation of ACF. As a result of windowing, the ACF tapers as the autocorrelation lags increase. Boersma demonstrated that the tapering effect could be compensated for by dividing the ACF of the windowed signal by the autocorrelation of the windowing function itself, referred to as wACF hereafter. We recently found that wACF could cause overcompensation and therefore, result in errors in pitch detection. In this paper, a novel normalization method, eACF, is proposed that can both mitigate the tapering effect and minimize the overcompensation. The new method is evaluated on synthetic speech and on the TIMIT database with various types of additive noise at different signal-to-noise (SNR) ratios. The results show that the new method leads to better performance both in terms of pitch detection and speech activity detection. In this paper, we also investigate the scenarios where applying the wACF method is advantageous and where it is not.

AM Modulation Signal Estimation Allowing Further Research on Sources of Voltage Fluctuations

“…For each component signal obtained from the decomposition process, the fundamental frequency f mic of i-th component signal and its amplitude k mic in the considered 1 min measurement window were determined using particular methods of decomposition. To determine the frequency f of i-th nent signal obtained from decomposition methods other than DAPW, the autocorrelation function was used with a window function that allows determining the fundamental frequency of any function [31]. In this research, the window function was assumed to be a moving average function.…”

Harmonic Differences Method for Robust Fundamental Frequency Detection in Wideband and Narrowband Speech Signals

“…e fierce debate on missing fundamental led to the calculation of fundamental frequency, and many algorithms flourished to extract and track the pitch of a signal. Among the most notables are AMDF [18][19][20], autocorrelation [13][14][15][16][21][22][23][24][25], cepstrum [26][27][28], harmonic product spectrum, period histograms [29][30][31], parallel processing methods [32][33][34], simplified inverse filter tracking (SIFT) [35], comb filters [36], data reduction [37], LPC-based spectral equalization (unpublished), spectral sieves [38], harmonic spacings and structures [39][40][41], LPC inverse filtering [20], feature based [42], IPTA [43], harmonic pattern recognition [44], envelop analysis, threshold-crossing analysis (ZXABE, TABE, TTABE) [45,46], subharmonic summation [47], subband processing [48], superresolution [49], two-way mismatch [50], resolution improvement [51], TEMPO [52], RAPT (NCCF) [53], instantaneous frequency [54][55]…”

Robust Period Estimation of Automated Cutting Systems by Improved Autocorrelation & Linear Regression Techniques

“…Another drawback of the AC is that the lag which corresponds to the biggest peak is not necessarily the period of the signal, but rather an integer multiple of it, especially when the signal is quasi-periodic as is the case here. This phenomenon is known as "octave error" [5]. Also, for small values of the lag τ , the correlation will be significantly higher than the peak where τ T , which means that finding small periods becomes less reliable.…”

“…with X and Y the x and y coordinates of the maxima positions. The estimated period is corrected by the slope m as defined in (5).…”