Speech coding has evolved into a highly essential branch of applied signal processing, resulting in an explosion of products such as cellular, answering machines, mobile telephony and most recently voice-over-IP. Because of the increasing demand for speech communication. Speech coding technology has captured more and more attention from the research, standardization and business communities. All of whom have been encouraged to pursue and investigate alternative coding methods with the objective of minimizing its limitation and deficiencies.

Objective:
This course is specially designed to cover the basic speech coding methods in the modern-day speech coder. Numerous hands-on Matlab-based exercises are also included as part of the course to help participants enrich and reinforce the learning process.

Prerequisite:
A keen interest in audio/speech processing and a desire to acquire better conceptual understanding of this subject is desirable. Technical background knowledge in digital signal processing is also helpful.

Duration:
2 Full Days

Course outline

Day 1

1. Speech production model
2. Linear Prediction Coding (LPC) model of speech production

1. Magnitude sum function
2. Zero-crossing rate
3. The voicing classification problem

1. Autocorrelation method
2. Magnitude difference function
3. Multiplicity check

Fractional pitch period estimation

Day 2

Linear Prediction Coding Analysis:

1. Spectrum estimation
2. Prediction gain
3. The Levinson-Durbin Algorithm
4. Long-Term Linear Prediction

1. LPC encoder
2. LPC decoder

The Code-Excited Linear Prediction (CELP) speech coding system:

1. The Analysis-by-Synthesis Speech Coding Approach
2. CELP encoder and decoder

Who should attend?

The primary target audiences for this course are researchers and engineers with little or no background in speech coding systems but are keen to learn more about this subject. The course will also be of interest to technical managers who supervise research or development related to speech coding applications.