Search

Alternative features were derived from extracted temporal envelope bank (TBANK). These simplified temporal representations were investigated in alignment procedures to generate frame-level training labels for deep neural networks (DNNs). TBANK features improved temporal alignments both for supervised training and for context dependent tree building.

Traditionally, only five components are regarded as having to do with the special characteristics of recognizing tones, while front-end processing and feature extraction have been considered essentially independent. Since mismatch between training and testing in signal-space leads to subsequent distortions in feature-space and model-space, determining whether front-end processing and feature extraction is independent or dependent will be critical for robustness.

Anchorperson segment detection enables efficient video content indexing for information retrieval. Anchorperson detection based on audio analysis has gained popularity due to lower computational complexity and satisfactory performance. This paper presents a robust framework using a hybrid I-vector and deep neural network (DNN) system to perform anchorperson detection based on audio streams of video content. The proposed system first applies I-vector to extract speaker identity features from the audio data. With the extracted speaker identity features, a DNN classifier is then used to verify the claimed anchorperson identity. In addition, subspace feature normalization (SFN) is incorporated into the hybrid system for robust feature extraction to compensate the audio mismatch issues caused by recording devices. An anchorperson verification experiment was conducted to evaluate the equal error rate (EER) of the proposed hybrid system. Experimental results demonstrate that the proposed system outperforms the state-of-the-art hybrid I-vector and support vector machine (SVM) system. Moreover, the proposed system was further enhanced by integrating SFN to effectively compensate the audio mismatch issues in anchorperson detection tasks.

This study proposes a polynomial based feature transferring (PFT) algorithm for acoustic feature conversion. The PFT process consists of estimation and conversion phases. The estimation phase aims to compute a polynomial based transfer function using only a small set of parallel source and target features. With the estimated transfer function, the conversion phase converts large sets of source features to target ones. This study evaluates the proposed PFT algorithm using a robust automatic speech recognition (ASR) task on the Aurora-2 database. The source features were MFCCs with cepstral mean and variance normalization (CMVN), and the target features were advanced front end features (AFE). Compared to CMVN, AFE provides better robust speech recognition performance but requires more complicated and expensive cost for feature extraction. By PFT, we intend to use a simple transfer function to obtain AFE-like acoustic features from the source CMVN features. Experimental results on Aurora-2 demonstrate that the PFT generated AFE-like features that can notably improve the CMVN performance and approach results achieved by AFE. Furthermore, the recognition accuracy of PFT was better than that of histogram equalization (HEQ) and polynomial based histogram equalization (PHEQ). The results confirm the effectiveness of PFT with just a few sets of parallel features.