Tone and speech recognition in communications systems

Abstract

The present invention provides an apparatus and method for verifying address identifiers over a telecommunications network by first storing a plurality of zip codes in a database. A corresponding list of address identifiers is created and stored for each zip code. After receiving a spoken zip code from a user, the corresponding list of address identifiers is retrieved. Once the list of address identifiers is retrieved, a plurality of choices for at least some of the address identifiers is created. These plurality of choices contain words that rhyme with or sound like other address identifiers located in the retrieval group of address identifiers. After the system receives a spoken address identifier from the user, the system queries the database for a match between the spoken address identifier and the stored address identifier. The user is then prompted to verify that the system accurately recognized the spoken address identifier. However, if a plurality of choices was created for the matched address identifier, the user is presented with the plurality of choices to select from instead of the matched address identifier. Whether the address identifier or the plurality of choices is presented to the user, the user is requested to verify that the system correctly recognized the spoken address identifier.

What is claimed is:  
     
         1 . A method of recognizing speech data comprising the steps of 
 prompting a user to provide at least one set of numeric data via a keypad of a communications device;    querying a database to match said at least one set of numeric data to one or more entries in said database;    retrieving a pre-stored set of sequence of speech patterns using said at least one set of numeric data as a search key for such retrieving;    prompting said user to provide speech data via a mouthpiece of said communications device; and    in response to receiving said speech data, comparing said speech data to said sequence of speech patterns to recognize said speech data.    
     
     
         2 . The method of    claim 1    wherein said at least one set of numeric data is a zip code.  
     
     
         3 . The method of    claim 1    wherein said speech data is a home address.  
     
     
         4 . The method of    claim 1    wherein said at least one set of numeric data is a zip code and said speech data is a home address  
     
     
         5 . The method of    claim 1    wherein said at least one set of numeric data includes a zip code and a street number.  
     
     
         6 . A method of recognizing speech data comprising the steps of 
 prompting a user to enter a first set of numeric data via a keypad of a communications device;    querying a database to match said first set of numeric data to an entry in said database;    prompting said user to enter a second set of numeric data via said keypad of said communications device;    comparing said second set of numeric data to pre-stored range of digits;    retrieving from said second database a pre-stored set of sequence of speech patterns using said first set of numeric data as a search key for such retrieving;    prompting said user to provide speech data via said communications device; and    in response to receiving said speech data from said user, comparing said speech data to said sequence of speech patterns using said second set of numeric data as input to recognize said speech data.    
     
     
         7 . The method of    claim 6    wherein said second set of numeric data is also used to retrieve said sequence of speech patterns.  
     
     
         8 . The method of    claim 6    wherein said speech data is a home address.  
     
     
         9 . The method of    claim 1    wherein said first set of numeric data is a zip code and said speech data is a home address.  
     
     
         10 . The method of    claim 1    wherein said first set of numeric data includes a zip code and said second set of numeric data is street number.  
     
     
         11 . A method of recognizing speech data comprising the steps of: 
 prompting a user to provide one or more sets of Dual Tone Multi Frequency (DTMF) data via a keypad of a communications device;    prompting said user to provide speech data via said communications device;    recognizing said speech data using said DTMF data as input to facilitate said speech recognition.    
     
     
         12 . The method of    claim 11    wherein said one or more sets of numeric data is selected from a group which includes a zip code and a street number and wherein said speech data is a street name.  
     
     
         13 . A method for verifying an address identifier over a network comprising the steps of: 
 receiving a zip code provided by a user;    querying a database of stored zip codes wherein each stored zip code is indexed to a group of address identifiers;    determining whether said received zip code matches one of said stored zip codes;    if one of said stored zip codes matches the received zip code, retrieving said indexed group of address identifiers corresponding to said matched zip code;    receiving a speech signal from said user; and    recognizing said speech signal by verifying that said speech signal matches one of said address identifiers.    
     
     
         14 . The method according to    claim 13   , wherein said address identifiers include names of cities and towns.  
     
     
         15 . The method according to    claim 13   , wherein said address identifiers include names of states.  
     
     
         16 . The method according to    claim 13   , wherein said address identifiers are names of streets.  
     
     
         17 . The method according to    claim 16   , wherein said names of streets further include street numbers.  
     
     
         18 . The method according to    claim 13    wherein said recognizing step comprises the steps of: 
 gathering from said speech signal information to generate a plurality of likely addresses;  
 filtering said likely addresses to derive a limited number of choices of likely addresses; and  
 presenting said limited number of choices of likely addresses to said user for verification.  
 
     
     
         19 . The method of    claim 18    further including the step of: 
 receiving a signal from said user indicative of a selection by said user of one of said likely addresses; and  
 storing said user-selected address in a storage area.  
 
     
     
         20 . The method of    claim 18    wherein said filtering step comprises the step of 
 screening out one or more likely addresses for which street numbers as recognized in said speech signal, are outside of a predetermined range for street names associated with said one or more likely addresses.  
 
     
     
         21 . The method according to    claim 13   , wherein said received zip code is provided by said user as speech data.  
     
     
         22 . The method according to    claim 13   , wherein received zip code is provided by said user as dual tone multi frequency (DTMF) signal from a keypad.  
     
     
         23 . A speech recognition system for recognizing address identifiers comprising: 
 a first database for storing zip codes; 
 a second database for storing address identifiers indexed by said zip codes; and  
 a speech recognizer which a) is coupled to said first and second databases and which retrieves address identifiers from said second database in response to receiving a user-provided zip code and b) includes a software-driven processor that recognizes speech signals received from a user.  
   
     
     
         24 . The speech recognition system of    claim 23   , further comprising means for prompting said user to provide data to said speech recognition system via a keypad of a communications device.  
     
     
         25 . The speech recognition system of    claim 23   , further comprising means for prompting said user to provide data to said speech recognition system via a microphone of a communications device.