Module 4 - The Human Voice

Voice  ::  Speech Production  ::  Speech Recognition  ::  Singing

Speech Recognition

For years, engineers and scientists have worked to build machines that will recognize speech. This technology is getting better. There is a company in Utah called Fonix, developing sophisticated speech recognition software that does not require you to "train" the computer with your voice before it will work. However this speech recognition technology only works with a limited vocabulary, and is still far behind what the human speech recognition system can do.

According to the text book, humans can follow speech at rates up to 400 words a minutes with 30 individual sounds, we call them phonemes, per second. Even everyday conversational speech "requires the recognition of 10 - 15 phonemes per second.

In this chapter you will learn about the tools that can be used to analyze speech.

To successfully complete this chapter you should...

• Read chapter 16 in your text book (pages 329- 345).
• Carefully study the important concepts about speech recognition, analysis, and synthesis as you read the chapter.
• Be able to to list several levels of speech recognition.
• Be able to list the different tools that are used for speech analysis.
• Be able to describe several methods for speech synthesis.

Important Concepts

Analysis of speech
Oscilloscope
Spectrum analyzer
Spectrograph
3-D plot

Speech recognition
Recognition of vowels
Recognition of consonants
Filtered speech & noisy environments
Coarticulation
Voiceprints

Synthesis of speech
Formant synthesis (vocoder)
LPC
Sampling