Speech recognition system and method permitting user customization

• US 6,873,951 B1
• Filed: 09/29/2000
• Issued: 03/29/2005
• Est. Priority Date: 03/30/1999
• Status: Expired due to Fees

• Current Assignee: RPX Clearinghouse LLC (RPX Corporation)
• Original Assignee: Nortel Networks Limited (Nortel Networks Corporation)
• Inventors: Lin, Ping, FLORENCE, LLOYD, Adelman, Jason, Lin, Lin
• Primary Examiner(s): Dorvil, Richemond
• Assistant Examiner(s): Storm, Donald L.
• Application Number:
  US09/672,814
• Time in Patent Office: 1,642 Days
• Field of Search: 704/244, 704/251, 704/231, 704/275, 704/243, 379/88.01, 379/88.03
• US Class Current: 704/251
• CPC Class Codes: G10L 15/065 Adaptation

Abstract

A system and method for speech recognition includes a speaker-independent set of stored word representations derived from speech of many users deemed to be typical speakers and for use by all users, and may further include speaker-dependent sets of stored word representations specific to each user. The speaker-dependent sets may be used to store custom commands, so that a user may replace default commands to customize and simplify use of the system. Utterances from a user which match stored words in either set according to the ordering rules are reported as words.

Claims

1. A speech recognition system comprising computer memory storing:

a first set of speaker-independent word models used to match a word in an utterance of a user with a word model in said first set, wherein said first set of word models includes models for each of a plurality of words;

a second set of speaker dependent word models derived from speech of a particular user and used to match a word in an utterance of said particular user, wherein said second set of word models includes models for at least some of said plurality of words; and

a program portion used to identify words in utterances of said particular user by attempting to match portions of an audio signal with;

word models among said first set; and

word models among said second set,

wherein said identified words in the utterances of said particular user include user-selected words for invoking commands.

2. A method of operating a speech recognition system comprising:

storing a first set of speaker-independent word models used to match a word in an utterance of any user with a word model in said first set, said first set of word models including models for each of a plurality of words;

storing a second set of speaker dependent word models derived from speech of a particular user, said second set of word models including models for at least some of said plurality of words and at least one model of said second set chosen by said particular user to initiate performance of at least one of a plurality of system commands;

recognizing words in utterances of said particular user by attempting to match portions of an audio signal with;

word models among said first set; and

word models among said second set; and

performing at least one system command in response to a recognized word within said utterances of said particular user.

3. The method according to claim 2 further comprising:

inviting said particular user to speak training utterances of a word upon a predetermined number of failures to recognize said word using said first set of word models;

deriving a word model from said training utterances; and

storing said word model from said training utterances, in said second set.

4. A method of operating a speech recognition system comprising:

storing a first set of speaker-independent word models used to match a word in an utterance of any user with a word model in said first set;

storing a second set of speaker dependent word models derived from speech of a particular user by;

inviting said particular user upon first use of said speech recognition system to speak training words for deriving said second set;

deriving said second set from said training words; and

storing said second set;

associating at least one stored word model with a command token also associated with a default command word model; and

recognizing words in utterances of said particular user by attempting to match portions of an audio signal with;

word models among said first set; and

word models among said second set.

5. A method of operating a speech recognition system comprising:

storing a first set of speaker-independent word models used to match a word in an utterance of any user with a word model in said first set;

storing a second set of speaker dependent word models derived from speech of a particular user by;

determining a likelihood of recognizing a spoken word using said first set;

deriving a word model from a spoken word marginally recognized using said first set;

storing said word model in said second set; and

recognizing words in utterances of said particular user by attempting to match portions of an audio signal with;

word models among said first set; and

word models among said second set.

6. A method of enhancing speech recognition comprising:

providing a set of user-independent word models derived from utterances of a plurality of speakers, said set of user-independent word models including models for each of a plurality of words;

providing a set of user-dependent word models for ones of a plurality of users each derived from utterances of one of said users, said set of user-dependent word models including models for at least some of said plurality of words;

matching an utterance from one of said users to one of said user-independent word models;

matching an other utterance from said one of said users to one of said user-dependent word models; and

responsive to matching either said utterance to said one of said user-independent word models or said other utterance to said one of said user-dependent word models, initiating a command associated with both said one of said user-independent word models and said one of said user-dependent word models.

7. A method of enhancing speech recognition comprising:

providing a set of user-independent word models derived from utterances of a plurality of speakers, at least one user-independent word model representing a first word and associated with a command token;

providing a set of user-dependent word models for ones of a plurality of users each derived from utterances of one of said users, at least one user-dependent word model representing a second word different than the first word and associated with said command token, said user-dependent word models each derived by;

inviting a new user to speak training words for deriving a set of user-dependent word models;

deriving said set of user-dependent models from said training words;

storing said set of user-dependent word models; and

associating a user-dependent word model with a command token designated by said one of said users;

matching an utterance from one of said users to one of said user-independent word models; and

matching an other utterance from said one of said users to one of said user-dependent word models.

8. The method according to claim 7 further comprising:

inviting said new user to speak training utterances of a word upon a predetermined number of failures to identify said word among said user-independent word models when no model for said word is present in said user-dependent models;

deriving a word model from said training utterances; and

storing the derived word model in said set of user-dependent word models.

9. The method according to claim 8 wherein said user-dependent word models are stored in a separate memory location from said user-dependent word models.