CA461 Speech Processing 1

1
CA461 Speech Processing 1

• John McKenna

2
Introductory Lecture

• Welcome
• Admin
• Contact
• Prerequisites
• Assessment
• Module Overview
• Syllabus
• Learning Outcomes

3
Welcome

• CA4 students welcome from all streams
• CL4 core module
• CLX welcome too
• Please mail me if you have doubts about
  prerequisite knowledge

4
Contact Details

• Email
• john_at_computing.dcu.ie
• John.McKenna_at_computing.dcu.ie
• John.McKenna_at_dcu.ie
• Office
• Room L2.47
• Tel. [phone]

5
Logistics

• Lectures
• Twice a week
• Labs
• 1 x 2 hour lab per week (start Week 1)
• Moodle
• moodle.dcu.ie
• VLE
• Lecture notes, Discussion forums, etc

6
Prerequisites

• Open mind
• Some maths
• probability, linear algebra (matrices)
• Ability to program
• Problem solving skills
• Communication skills

7
Assessment

• Continuous Assessment 60
• 1 Assignment 50
• Issued about week 7 due week 12
• 4-page, conference-style paper on a
  speech/speaker recognition implementation
• APC 10
• End of module exam 40

8
APC

• Not a distance education module
• Attendance
• Performance
• Contribution

9
You will do well in this module if

• You think analytically
• Think for yourself
• Engage the subject
• Communicate well

10
Materials

• Books
• See Module Descriptor for list
• No book purchase necessary
• Recommended
• Gold Morgan, or Holmes Holmes
• Headset required
• Composite (with microphone) recommended
• Sharing feasible

11
Indicative Syllabus

• General
• To present the characteristics of speech
• To discuss automatic speech recognition systems
• Specific
• Speech Production, Representations and
  Terminology
• Acoustic Phonetics
• Overview of ASR (Automatic Speech Recognition)
• Speech Parameterisation for ASR
• HMMs and Trellis Algorithms
• HMM Recognition and Training
• Other issues and applications

12
Extensible Learning Outcomes

• Familiarity with the building blocks of language
• Understanding of time/frequency representations
  DSP
• Knowledge of pattern matching algorithms
• Ability to program MATLAB scripts
• Ability to use HTK (Hidden Markov Model Toolkit)
• Knowledge of the principles and problems in the
  design, implementation and evaluation of
  machine-learning systems

13
Speech Processing 2?

• Focus on
• Speech Analysis
• Speech Synthesis
• Prerequisites
• Speech Processing 1 or
• possibly DSP 1
• Semester 1
• You can choose both DSP1 and SP1

14
Next

• Try the first Lab
• Recording
• Transcription vs. Orthography
• Analysis
• Synthesis
• Next Lecture
• Sounds Speech Production