Analysis and improvement of multilingual speech coding

Mohamad Itani

Doctoral dissertation

Dissertations are not being sold



The thesis investigates the performance of voice codec’s that uses linear predictive coding (LPC), over different languages. The English language has had a dominating influence on the advance of telecommunications. With many of the major developments coming from primarily English speaking areas there is the risk that these advances may not be linguistically robust. It is noted that the quality of a speech produced by voice codecs mainly is assessed using samples of English language. Investigations show that most low-rate speech coders show bias towards non-accented English. When the coders are used for heavily accented English or other languages, significant performance degradation is noted.

The main goal of the thesis is to investigates and propose ways for improvement of the performance of CELP voice codecs for coding multilingual speech. In order to achieve the goal, the following problems had been solved: performed analytical review on aspects of multilingual CELP voice coding; created speech records of Lithuanian language database according to ITU-T Rec.

P.50 App. 1. that is suitable for tests of performance voice codecs; investigated performance of AMR and Speex codecs for coding non-English speech and proposed and proved ways for improvement of multilingual speech coding. Chapter 1 reviews most of the work done in the field of influence of language on coding, by introducing CELP based codes, and previous researches about the topic in hand. As well, it concludes in formulating of main objective and tasks of present investigation.

Chapter 2 introduces the quality measuring techniques categorized by subjective and objective. Moreover, it contains our primary experiment on the codecs and quality assessment techniques.

Chapter 3 presents the construction of the Lithuanian speech database. After concluding the results in chapter 2, we found ourselves in need of a well formed speech database for Lithuanian language to carry out our experiments on.

Chapter 4 presents the experimental study on the performance of CELP based codes under different languages.

Chapter 5 suggests a solution the codecs to perform better under different languages.

Three scientific articles have been published in reviewed scientific publications by the author within scope of research: two publications in reviewed ISI Web of Science journals and one in other reviewed scientific editions. Also results of performed investigations were presented in 4 scientific conferences.


Read electronic version of the book:


Book details

Data sheet

Imprint No:
102 p.
16 other books in the same category:

Follow us on Facebook