SPEECH RECOGNITION CIRCUIT USING PARALLEL PROCESSORS

• US 20140365220A1
• Filed: 06/19/2014
• Published: 12/11/2014
• Est. Priority Date: 02/04/2002
• Status: Active Grant

• Current Assignee: Zentian Limited
• Original Assignee: Zentian Limited
• Inventors: Catchpole, Mark
• Granted Patent: US 9,536,516 B2
• Time in Patent Office: Days
• US Class Current: 704/251
• CPC Class Codes: G10L 15/05 Word boundary detection
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams
• CPC Class Codes: G10L 15/34 Adaptation of a single recogniser for parallel processing, e.g. by use of multiple processors or cloud computing

Abstract

A speech recognition circuit comprises an input buffer for receiving processed speech parameters. A lexical memory contains lexical data for word recognition. The lexical data comprises a plurality of lexical tree data structures. Each lexical tree data structure comprises a model of words having common prefix components. An initial component of each lexical tree structure is unique. A plurality of lexical tree processors are connected in parallel to the input buffer for processing the speech parameters in parallel to perform parallel lexical tree processing for word recognition by accessing the lexical data in the lexical memory. A results memory is connected to the lexical tree processors for storing processing results from the lexical tree processors and lexical tree identifiers to identify lexical trees to be processed by the lexical tree processors. A controller controls the lexical tree processors to process lexical trees identified in the results memory by performing parallel processing on a plurality of said lexical tree data structures.

Claims

1-91. -91. (canceled)

92. An apparatus comprising:

a receiving module for receiving a feature vector;

an updating module coupled to the receiving module for updating a model comprising triphones in response to the received feature vector, the updating module being configured to detect a word boundary in the model; and

a language model module coupled to the updating module, the language model module being activated in response to said detection of a word boundary to determine a language model score;

93. An apparatus according to claim 92, wherein the apparatus is capable of simultaneously processing different triphone models.

94. An apparatus according to claim 92, wherein the model is a Hidden Markov Model.

95. An apparatus according to claim 92, wherein the language model uses an N-gram model.

96. A method comprising:

receiving a feature vector;

updating a model comprising triphones in response to the received feature vector;

determining whether a word boundary has been reached in the model; and

if it has been determined that a word boundary has been reached, retrieving a score from a language model.

97. The method of claim 96, wherein the method operates within a hardware implemented speech recognition system capable of simultaneously processing different triphone models.

98. The method of claim 96, wherein the model is a Hidden Markov Model.

99. The method of claim 96, wherein the language model uses an N-gram model.