Voice Recognition

1
Voice Recognition

• Dennis Gehris
• Professor
• Bloomsburg University

2
Voice Recognition

• Voice Recognition Definitions
• Timeline of Voice Recognition Development
• Types of Voice Recognition Systems
• How Voice Recognition Works
• Voice Recognition Systems and Requirements
• Voice Recognition Demonstration

2
3
Voice Recognition Definitions

• The field of computer science that deals with
  designing computer systems that can recognize
  spoken words.
• Speech-to-Text
• The field of computerized voice recognition /
  speech recognition. (These terms are usually used
  interchangeably)

3
4
Voice Recognition Definitions (continued)

• Text-to-Speech
• Computerized enunciation of the written word. In
  a sense, this is the opposite of speech-to-text.
• Voice Identification, Voice Verification
• The use of voice recognition technology to
  identify a person by their voice, rather than to
  identify the meaning of the spoken word.

4
5
Voice Recognition Timeline

• Late 1950s
• voice recognition research begins.
• 1964
• IBM demonstrates Shoebox for spoken digits at New
  York World's Fair.

5
6
Voice Recognition Timeline (continued)

• 1968
• The HAL-9000 computer in 2001 A Space Odyssey
  introduces the world to voice recognition.
• 1978
• Texas Instruments introduces the first
  single-chip speech synthesizer and the Speak and
  Spell toy.

6
7
Voice Recognition Timeline (continued)

• 1993
• IBM launches the first packaged voice recognition
  product, the IBM Personal Dictation System for
  OS/2.
• 1993
• Apple ships PlainTalk, a series of speech
  recognition and speech synthesis extensions for
  the Macintosh.

7
8
Voice Recognition Timeline (continued)

• 1994
• Dragon Systems' DragonDictate for Windows 1.0 is
  the first software-only PC-based dictation
  product.
• 1996
• IBM introduces MedSpeak/Radiology, the first
  real-time continuous-speech recognition product.

8
9
Voice Recognition Timeline (continued)

• 1996
• OS/2 Warp 4 becomes the first operating system to
  include built-in voice navigation and
  recognition.
• June 1997
• Dragon ships NaturallySpeaking, the first
  general-purpose continuous-voice recognition
  product.

9
10
Voice Recognition Timeline (continued)

• August 1997
• IBM ships ViaVoice.
• Fall 1997
• Microsoft CEO Bill Gates identifies voice
  recognition as a key technological advance.

10
11
Types of Voice Recognition Systems

• Discrete Speech Systems
• require that the speaker speak slowly and
  distinctly and separate each word with a short
  pause
• The requirement to pause, usually between 1/10th
  and 2/10ths second between words
• Continuous Speech Systems
• allow you to speak naturally without pauses
  between the words

11
12
Types of Voice Recognition Systems (continued)

• Most systems are Speaker Dependent
• requires an enrollment (training) period
  (usually 15 - 60 minutes) in order to understand
  a new speaker
• Some systems are Speaker Independent
• does not need an enrollment period to start
  recognition. A system which will comprehend any
  person who speaks into the system as soon as they
  start speaking

12
13
How Voice Recognition Works

• Speech Engine --
• The key to speech recognition
• The mechanism that translates sounds into words
  and sentences.

14
How Voice Recognition Works (continued)

• 1. Audio Input
• The speech engine receives the audio input, or
  speech, from a microphone via a PC with a
  standard sound card such as the Creative Labs
  Sound Blaster.

15
How Voice Recognition Works (continued)

• 2. Acoustic Processor
• The acoustic processor, filters out background
  noise and converts the capture audio into a
  series of sounds that correspond to the
  phonemes--units of speech--making up the selected
  language, such as American English.

16
How Voice Recognition Works (continued)

• 3. Word Matching
• The speech engine attempts to match the sounds to
  the most likely words in two basic ways.

17
How Voice Recognition Works (continued)

• 3. Word Matching (continued)
• a. acoustical analysis is used to build a list
  of possible words that contain similar sounds.
• b. language modeling is used (the likelihood
  that a given word would appear between those
  coming before and after it) to narrow the list
  and come up with the best candidates.

18
How Voice Recognition Works (continued)

• 4. Decoder
• The decoder selects the most likely word based on
  the rankings assigned during word matching and
  assembles the words in the most likely sentence
  combinations.

19
How Voice Recognition Works (continued)

• 5. The End Result
• Results sometimes appear in the application and
  are saved as an application file
• Dragon Systems' Dragon NaturallySpeaking and
  IBM's ViaVoice include their own word processors.
  Results are saved in rich-text format or pasted
  it into another application.

20
Voice Recognition Demonstration