Data processing: speech signal processing – linguistics – language – Speech signal processing – Application
Reexamination Certificate
2000-06-21
2004-05-25
Abebe, Daniel (Department: 2655)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Application
C704S246000
Reexamination Certificate
active
06741963
ABSTRACT:
CROSS REFERENCE TO RELATED APPLICATIONS
(Not Applicable)
STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT
(Not Applicable)
BACKGROUND OF THE INVENTION
1. Technical Field
This invention relates to the field of speech recognition software, and more particularly, to using a speech cache in conjunction with a speech recognition `application to improve system performance.
2. Description of the Related Art
Speech recognition is the process by which an acoustic signal received by microphone is converted to a set of text words by a computer. These recognized words may then be used in a variety of computer software applications for purposes such as document preparation, data entry, and command and control. Improvements to speech dictation systems provide an important way to enhance user productivity.
Some speech recognition applications cannot recognize a user spoken utterance identifying a word or a word phrase without the aid of attributes. This is particularly true of embedded speech recognition applications having limited vocabularies, such as the variety used in navigation systems in automobiles. Attributes provide the speech recognition system with supplemental information detailing the user spoken utterance. Oftentimes, for such a speech recognition system to recognize a user spoken utterance, the user must also issue a lengthy series of attributes. For example, if the user utters a phrase recognized as a speech command such as “how far” in conjunction with a speech object, “Roller Coaster World”, then the system may require attributes identifying the object within the speech command. In this case, to properly identify the object “Roller Coaster World” to the speech recognition system, attributes such as “U.S.A.”, “Florida”, “Orlando”, and “amusement park named Roller Coaster World” may be necessary. Using attributes to specify a speech object within a speech command can be analogized to navigating through a system of computer directories to find a particular computer file.
Even more troublesome is the case when the user repeatedly issues the same speech command or issues a series of subsequent speech commands involving the same object. This situation commonly occurs in the case of a user driving to a distant location for vacation. Conventional systems do not store previously issued speech commands, objects, or attributes. Thus, each time the user issues a speech command regarding a previously identified object, the user must also provide the system with the previously mentioned attributes identifying the object. For example, each user command requesting information such as the distance or route to “Roller Coaster World” must be accompanied by the lengthy list of attributes identifying “Roller Coaster World” to the system. A significant amount of time and efficiency could be gained if speech recognition systems having a limited vocabulary could more efficiently recall previously used commands or objects.
SUMMARY OF THE INVENTION
The invention disclosed herein for improving system performance of speech systems in accordance with the inventive arrangements satisfies the long-felt need of the prior art by using a speech cache and speech cache logic in conjunction with the recognition system. Such speech systems can recall particular objects or speech commands from a speech cache, thereby eliminating the need for users to continually utter redundant attributes to the speech system in an effort to properly describe a speech object. Because speech systems frequently use a set of commands or objects, the cache is a cost effective method of enhancing memory systems using statistical means, without having to resort to the expense of making the whole memory system faster.
The invention concerns a method and a system for improving recall of speech data in a computer speech system. Significantly, the speech system can be an embedded computer speech system. The method of the invention involves a plurality of speech cache management steps including providing a speech cache; receiving a speech system input and identifying a speech event in the received speech system input, the speech event comprising speech data. Subsequently, the speech data can be compared to pre-determined speech cache entry criteria; and, if the speech data meets one of the pre-determined entry criteria, at least one entry can be added to the speech cache, the at least one entry corresponding to the speech data. Additionally, the speech data can be compared to pre-determined speech cache exit criteria; and, if the speech data meets one of the pre-determined exit criteria, at least one entry can be purged from the speech cache, the at least one entry corresponding to the speech data.
In the preferred embodiment, the entry criteria comprises frequently used speech data, recently used speech data, and important speech data. Similarly, the exit criteria can comprise least frequently used speech data associated with each entry in the speech cache, least recently used speech data associated with each entry in the speech cache and least important speed data associated with each entry in the speech cache.
The method of the invention can also include a speech cache filtering process. Specifically, an embodiment incorporating speech cache filtering can compare entries in the speech cache with filtering criteria; and, sort the entries according to the filtering criteria. The filtering criteria can comprise frequency of use of speech data associated with each entry in the speech cache, least recency of use of speech data associated with each entry in the speech cache, and importance of use of speech data associated with each entry in the speech cache.
In the preferred embodiment, the speech system input can be one of a system event and a speech event. To accommodate system events, the method of the invention can further establish a table of system events and corresponding speech cache commands. Responsive to receiving a system event, the received system event can be compared to the system events in the table. If the received system event matches a system event in the table, the speech cache command corresponding to the matching system event in the table can be performed. Notably, the corresponding speech cache commands can include purge commands and add commands.
In the preferred embodiment, the comparing step can comprise evaluating the speech system input against user-configurable rules for adding and deleting from the speech cache entries corresponding to the speech data, the rules based on frequency of use of the speech data, recency of use of the speech data and importance of use of the speech data. Similarly, the comparing step can comprise the step of evaluating the speech system input against system configured rules for adding and deleting from the speech cache entries corresponding to the speech data, the rules based on a pre-specified list of speech data. Significantly, comparisons performed against the system-specified entry and exit criteria can be overridden with the comparisons performed against the user-specified entry and exit criteria
In a preferred embodiment, the method of the invention can further include establishing a frequency counter for the speech data. Responsive to receiving a speech event, the frequency counter corresponding to the speech data can be incremented. Thus, the comparing step can comprise evaluating the speech system input against user-configurable rules for adding and deleting from the speech cache entries corresponding to the speech data, the rules based on frequency of use of the speech data. In that case, the frequency can be measured by the frequency counter established for the speech data. Moreover, the adding step cans be performed in response to a frequency indicated by the frequency counter exceeding a pre-determined threshold. Likewise, the deleting step can be performed in response to a frequency indicated by the frequency counter falling below a pre-determined threshold.
Advantageously, the method can further comprise the steps of sensing when the speech cache is full
Badt Daniel E.
Guasti Peter J.
Hanson Gary R.
Nassiff Amado
Rodriguez Edwin A.
Abebe Daniel
Akerman & Senterfitt
LandOfFree
Method of managing a speech cache does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method of managing a speech cache, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method of managing a speech cache will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3225396