Speech Processing by Computer

LECTURE 3

DIGITAL FILTERING

Objectives

By the end of the session you should:

• understand that the basic building blocks of digital filters are poles and zeros

• understand that poles form resonators defined by their resonant frequency and bandwidth

• understand that complex systems can be constructed by cascading simple systems

• be able to explain how a low-pass filter can be designed by careful selection of resonators

• have a qualitative understanding of how band-pass and high-pass filters can be constructed along similar lines to low-pass filters

• be able to identify applications for filterbanks

• understand how a filterbank can be constructed to deliver estimates of energy over a range of frequencies.

• understand how a filterbank can also be used to synthesize speech

Outline

3. Digital Filtering

3.1. System modelling with poles and zeros

3.1.1. Non-recursive digital systems are all-zero systems

3.1.2. Purely recursive digital systems are all-pole systems

3.1.3. Positioning poles and zeros on response plot

3.2. Simple filter designs

3.2.1. Low-pass filter designs

3.2.2. High-pass and band-pass transformations

3.3. Filterbank analysis/synthesis

3.3.1. Bank of bandpass filters

3.3.2. Rectification

3.3.3. Smoothing

3.3.4. Source parameters

3.3.5. Re-synthesis

Reading

MATLAB Users:

MATLAB Signal Processing Toolbox Help: Filter Requirements and Specification

General:

C. Speaks, Introduction to Sound, Singular Press, 1996, Chapter 6. , Resonance and Filtering