System for speech recognition

A sound recognition system comprising:

• a microphone;
• a processor for digitizing sound signals received from the microphone;
• a memory coupled to the processor for storing pattern character strings;
• a comparator for comparing a sound character string formed from the sound signals to the pattern character strings, during a first comparison, the comparator selecting an N number of characters from the sound character string and comparing the N number of characters to each of the pattern character strings N times, and during a second comparison, the comparator comparing remaining characters of the sound character string to remaining characters of the each pattern character string M times where the remaining characters of both the sound character string and the each pattern character string are not compared during the first comparison;
• and
• a controller which takes a result of the first and the second comparisons from the comparator and calculates a first match probability based on the first comparison and a second match probability based on the first and the second comparisons, where the controller selects a group of pattern character strings based on the first match probability and identifies a resultant pattern character string that recognizes the sound character string based on the second match probability.

A method for generating a character string recognized to represent speech, comprising:

• providing an input of a spoken character string;
• performing a first comparison which compares the input of the spoken character string to a set of pattern character strings and reduces the set of pattern character strings to a group of qualifying pattern character strings having a match probability that exceeds a predetermined threshold where the first comparison comprises comparing a portion of the spoken character string with the portion of each of the pattern character strings and determining a number of hits between the spoken character string and the each pattern character string;
• performing a second comparison which identifies a resultant pattern character string representative of a recognition of the spoken character string, where the second comparison comprises comparing a part of, or an entire remaining characters of the spoken character string to each pattern character string of the group of qualifying pattern character strings and determining a total number of hits between the spoken character string and one or more characters of the group of qualifying pattern character strings;
• generating the resultant pattern character string and processing the resultant pattern character string as the character string to represent the speech;
• and
• outputting the resultant pattern character string as the character string representing the speech for subsequent processing using the speech output.