A sub-band-based feature reconstruction approach for robust speaker recognition

Abstract

Although the field of automatic speaker or speech recognition has been extensively studied over the past decades, the lack of robustness has remained a major challenge. The missing data technique (MDT) is a promising approach. However, its performance depends on the correlation across frequency bands. This paper presents a new reconstruction method for feature enhancement based on the trait. In this paper, the degree of concentration across frequency bands is measured with principal component analysis (PCA). Through theoretical analysis and experimental results, it is found that the correlation of the feature vector extracted from the sub-band (SB) is much stronger than the ones extracted from the full-band (FB). Thus, rather than dealing with the spectral features as a whole, this paper splits full-band into sub-bands and then individually reconstructs spectral features extracted from each SB based on MDT. At the end, those constructed features from all sub-bands will be recombined to yield the conventional mel-frequency cepstral coefficient (MFCC) for recognition experiments. The 2-sub-band reconstruction approach is evaluated in speaker recognition system. The results show that the proposed approach outperforms full-band reconstruction in terms of recognition performance in all noise conditions. Finally, we particularly discuss the optimal selection of frequency division ways for the recognition task. When FB is divided into much more sub-bands, some of the correlations across frequency channels are lost. Consequently, efficient division ways need to be investigated to perform further recognition performance.

A Review on Feature Extraction for Speaker Recognition under Degraded Conditions

“…Features are reconstructed from associated sub-bands, and then recombined. Compared with the full-band method [20], the proposed sub-bands method [31] showed a superior performance. The performances of these systems are mostly dependent on the accuracy of the noise estimation.…”

“…In [20], an IBM is constructed by estimating local SNR of elements in the time-frequency matrix representation of the speech. Instead of reconstruction, the noisy features using the whole spectrum, sub-bands are considered in [31]. Features are reconstructed from associated sub-bands, and then recombined.…”

“…However, it should be noted that power spectral subtraction is used in the preprocessing to suppress additive noise, which probably increased the recognition rate for both the baseline methods, and the proposed method. Two-dimensional (2D) [19À20], [23], [31], [34], [38], [47], [49], [51], [54À58] auto-regressive (AR) spectrograms are used in [43] to eliminate noise effects. In this method, two LP analysis, one in time domain, and the other in frequency domain, are applied to speech signal.…”

A robust polynomial regression-based voice activity detector for speaker verification

“…Similar to the VAD, vowellike regions are used in [29] and improved in [30] by including the non-vowel-like regions. Missing data approach is also investigated in several studies [31][32][33][34], where a binary time-frequency mask is constructed for the noisy spectrum to indicate reliable and unreliable features. The unreliable features are then reconstructed, or marginalized (ignored in score computation).…”