Effective Pitch Estimation using Canonical Correlation Analysis
Subrata Kumer Paul, Rakhi Rani Paul · 2020
Effective pitch estimation is very challenging task in recent research area. There are a large set of methods that have been developed in the speech processing area for the estimation of pitch. In this paper, focus on an effective pitch estimation method using Canonical Correlation Analysis (CCA). It is used for predicting IMF components from reference signals. The proposed pitch estimation algorithm dominant harmonic model enhanced the pitch peak through signal reshaping, then for on-linear and non-stationary data analysis method EMD uses to decompose the signal into a finite number of band limited signal called IMFs. Each IMF have specific frequency band. The Keele pitch reference database is considered for evaluation of the proposed method. Autocorrelation function to detect pitch period. Finally, Multivariate multidimensional statistics data analysis tools CCA is used for selecting IMF components. Then summing up IMF components for a partial reconstruction of the NACF signal except for noisy components. The reconstructed signal contains less noise. We estimated pitch from the reconstructed signal. The performance of the proposed pitch estimation is compared in terms of Gross Pitch Error (% GPE) with existing algorithms. Pitch is an important speech parameter for a wide range of applications such as speech coding, text to speech synthesis, speech recognition, gender identification, accent identification, accent synthesis, speaker identification, speaker verification, and so on.