Enhancement of substitution voices using F1 formant deviation analysis and DTW based template matching

Title

Enhancement of substitution voices using F1 formant deviation analysis and DTW based template matching

Creator

Inbanila K.; Krishnakumar E.

Description

Speech is the best way to express the thoughts and feelings among the human beings. But for many reasons the sound produced by human beings becomes disordered voice and termed with many names based on the cause as stammering, dys-theria, apraxia and so on. In the above mentioned few examples, the voice becomes disordered because of the underperformance of body's organ. The larynx is removed in some human beings because of cancer. For them an artificial larynx transducer (ALT) is used to produce the sounds. The above all sounds are categorized as disordered voice and the sound produced by ALT is also called as Substitution voice. In this paper, a method is used to improve the quality of substitution voice produced by ALT. Algorithm is developed to estimate undesired audio components from the device output and remove the same using Non Linear Spectral Subtraction (NLSS) technique. Further, Fundamental (F0) contour and novel parameter F1 formant deviation of healthy speech (HE) and ALT speech are determined. The above two parameters are estimated and stored during the training phase of the system. In the test phase, the above mentioned parameters are estimated and they are used to scale down the database to reduce overall enhancement time. Next step is template matching done by mapping test data with training data using Dynamic Time Warping (DTW) Technique. The data base with least distance estimation is recognized as the utterance and the same is played back. 2017 IEEE.

Source

Proceedings of the 2017 International Conference on Wireless Communications, Signal Processing and Networking, WiSPNET 2017, Vol-2018-January, pp. 352-356.

Date

2017-01-01

Publisher

Institute of Electrical and Electronics Engineers Inc.

Subject

ALT speech; DTW; F0 contour; F1 Formant deviation; Healthy Speech; Linear Prediction Cepstral coefficients (LPCC)

Coverage

Inbanila K., Department of ECE, Faculty of Engineering, Christ University, Bangalore, India; Krishnakumar E., Department of ECE, Visvesvarya Technological University, East Point College of Engineering and Technology, Bangalore, India

Rights

Restricted Access

Relation

ISBN: 978-150904441-2

Format

Online

Language

English

Type

Conference paper

Identifier

https://doi.org/10.1109/WiSPNET.2017.8299777

https://www.scopus.com/inward/record.uri?eid=2-s2.0-85046483637&doi=10.1109%2fWiSPNET.2017.8299777&partnerID=40&md5=dc929274f9a5e07b6929035ef26df738