Noise and Echo Aware Accurate Dysarthria Speech Recognition Model
Usha.M1, L. Sankari2
1Usha.M, Research Scholar, Sri Ramakrishna College of Arts and Science for Women, Coimbatore, Tamil Nadu, India.
2Dr .L. Sankar Associate Professor, Department of Computer Science, Sri Ramakrishna College of Arts and Science for Women, Coimbatore, Tamil Nadu India.
Manuscript received on November 20, 2019. | Revised Manuscript received on November 28, 2019. | Manuscript published on 30 November, 2019. | PP: 7531-7536 | Volume-8 Issue-4, November 2019. | Retrieval Number: D5333118419/2019©BEIESP | DOI: 10.35940/ijrte.D5333.118419
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Dysarthria speech is the speech disorder which would be caused due to weakness of human muscles. The human with dysarthria disorder cannot speak normally whose speech will be very slow and congestive words which might be more difficult to understand. Thus it is required to create the environment where the speech of dysarthria disorder people can be recognized. This is done in our previous research work by introducing the method namely Hidden Markov Model based Speech Recognition (HMMSR). However this research work didn’t focus on accurate prediction and the echo noises presence in the speech signals. This would lead to inaccuracy in speech recognition. This is resolved by introducing the efficient dysarthria speech recognition framework namely Noise and Echo aware Dysarthria Speech Recognition Method (NE-DSRM). In this research work, Hybrid Least Mean Square-Adaptive Neuro Fuzzy Inference System (LMS-ANFIS) has been used for preprocessing. This method will remove both echo and noises present in the speech signals to ensure the accurate prediction outcome. And then speech recognition is performed by comparing the dysarthria speech with the phonological speech based on which relevancy would be identified. The accuracy of speech recognition can be improved by introducing the SVM based learning methodology which can classify the dysarthria speech based on which more relevant matching can be done. The objective of the system is, after being trained, to identify and classify limited-vocabulary sets of speaker-dependent. The overall assessment of the research work is done in the matlab simulation environment from which it is proved that the proposed method NE-DSRM tends to have better performance than the existing research works.
Keywords: Dysarthria Speech, Phonological Speech, Noises, Echoes, Vocabulary Classification, Accurate Speech Recognition.
Scope of the Article: Classification.