Articulation-Disordered Speech Recognition Using Speaker-Adaptive Acoustic Models and Personalized Articulation Patterns

Abstract

This article presents a novel approach to speaker-adaptive recognition of speech from articulationdisordered speakers without a large amount of adaptation data. An unsupervised, incremental adaptation method is adopted for personalized model adaptation based on the recognized syllables with high recognition confidence from an automatic speech recognition (ASR) system. For articulation pattern discovery, the manually transcribed syllables and the corresponding recognized syllables are associated with each other using articulatory features. The Apriori algorithm is applied to discover the articulation patterns in the corpus, which are then used to construct a personalized pronunciation dictionary to improve the recognition accuracy of the ASR. The experimental results indicate that the proposed adaptation method achieves a syllable error rate reduction of 6.1%, outperforming the conventional adaptation methods that have a syllable error rate reduction of 3.8%. In addition, an average syllable error rate reduction of 5.04% is obtained for the ASR using the expanded pronunciation dictionary. -P. 2010. Articulation-disordered speech recognition using speakeradaptive acoustic models and personalized articulation patterns.

Diagnosing Dysarthria with Long Short-Term Memory Networks

“…The more advanced topic of recognizing speech produced from someone with Dysarthria using RNN networks has also been investigated recently for English speaking individuals, using Elman recurrent neural networks in [13] and a hybrid deep neural network -hidden Markov model (DNN-HMM) architecture in [14]. Wu et al [15] presented a personalized model adaptation for automatic speech recognition (ASR) targeted at Mandarin-speaking individuals afflicted with articulation disorders due to mild-to-moderate hearing impairment.…”

Pronunciation Adaptation For Disordered Speech Recognition Using State-Specific Vectors of Phone-Cluster Adaptive Training

“…Multiple pronunciations of a word in the lexicon improves the recognition performance. The lexical models are improved either implicitly or explicitly handling the pronunciation errors [13]. In order to improve the lexical models, the phones mispronounced by each dysarthric speaker need to be identified.…”

“…SLPAT 2015, 6th Workshop on Speech and Language Processing for Assistive Technologies, pages 72-78, Dresden, Germany, 11 September, 2015. c 2015 The Association for Computational Linguistics sonalized speaker articulation patterns were obtained from the speaker-adapted models along with the confusion matrix. These speaker-adapted models were obtained using universal disordered matrix and the posterior probability from the ASR system in an unsupervised fashion [13].…”

Recognition of Dysarthric Speech Using Voice Parameters for Speaker Adaptation and Multi-Taper Spectral Estimation

“…Studies [11] also suggest an improvement in recognition by using more suitable prior model or background model, for adaptation based on the dysarthric speaker's acoustic characteristics. Work pertaining to speaker based lexical or pronunciation model adaptation in addition to acoustic model adaptation, [12,13,14] have shown improvement in the ASR performance. An understanding of the speech production process through the articulatory models for speech has proven beneficial in improved accuracy of the ASR, both conventional GMM-HMM and DNN-HMM [1,15,16].…”