Acoustic Model Adaptation Using First-Order Linear Prediction for Reverberant Speech

Abstract

This paper describes a hands-free speech recognition technique based on acoustic model adaptation to reverberant speech. In handsfree speech recognition, the recognition accuracy is degraded by reverberation, since each segment of speech is affected by the reflection energy of the preceding segment. To compensate for the reflection signal we introduce a frame-by-frame adaptation method adding the reflection signal to the means of the acoustic model. The reflection signal is approximated by a first-order linear prediction from the observation signal at the preceding frame, and the linear prediction coefficient is estimated with a maximum likelihood method by using the EM algorithm, which maximizes the likelihood of the adaptation data. Its effectiveness is confirmed by word recognition experiments on reverberant speech.

A Bayesian view on acoustic model-based techniques for robust speech recognition

“…Subsections 4.3, 4.4, 4.6, and 6.9). Beyond this, however, the links of the decoding rules of the concepts of REMOS (Subsection 4.5), significance decoding (Subsection 4.7), modified imputation (Subsection 5.3), CMLLR/MLLR (Subsections 6.3 and 6.4), MAP (Subsection 6.5), Bayesian MLLR (Subsection 6.6), and Takiguchi et al [19] (Subsection 6.8) to the Bayesian framework via the mathematical reformulations in (28), (37), (45), (55), (61), (65), and (71), respectively, are explicitly stated for the first time in this paper.…”

“…In contrast to the approaches of Subsections 6.6.1 and 6.7, the concept proposed in [19] assumes the reverberant observation vector y n−1 at time n − 1 to be an approximation to the reverberation tail at time n in the logmelspec domain:…”

“…In this article, we review and examine for several uncertainty decoding [1][2][3][4][5], missing feature [6][7][8][9], and model adaptation techniques [10][11][12][13][14][15][16][17][18][19] how their compensation rules can be formulated as an approximated or modified Bayesian decoding rule. In order to illustrate the formalism, we also present the corresponding Bayesian network representations.…”

Model Adaptation for Long Convolutional Distortion by Maximum Likelihood Based State Filtering Approach

“…Feature-based methods like inverse-filtering and RASTA-PLP, despite their own limitations, have been more considerate of long convolutional distortion than most of the popular and successful model based approaches like PMC, VTS, universal adaptation and others, where compensation to convolutional distortion is generally based on a single state, without explicitly considering effect from preceding states or speech segments. Some of the model based approaches that explicitly consider the effect of preceding speech segments for adaptation include first-order linear prediction [12] and our previous work [13] based on state-splitting. In [12], energy component from preceding frames is estimated by first order linear prediction from (single) last frame of observation, and models are adapted at each frame, which is computationally very expensive and inefficient.…”

“…Some of the model based approaches that explicitly consider the effect of preceding speech segments for adaptation include first-order linear prediction [12] and our previous work [13] based on state-splitting. In [12], energy component from preceding frames is estimated by first order linear prediction from (single) last frame of observation, and models are adapted at each frame, which is computationally very expensive and inefficient. In our previous work [13], we proposed a state splitting approach to estimate preceding frames for a given state of HMM, which are used to compensate parameters of the state by convolving with channel parameters.…”