Data processing: speech signal processing – linguistics – language – Speech signal processing – Application
Reexamination Certificate
2000-04-12
2002-11-05
{haeck over (S)}mits, T{overscore (a)}livaldis Ivars (Department: 2641)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Application
C704S246000, C704S251000, C704S273000
Reexamination Certificate
active
06477500
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention generally relates to speech and speaker recognition and, more particularly, to text independent speaker recognition, which may thus include consistency checks of the recognition process, particularly suitable for continuous security and access control and which is transparent to one or more users and provides enhanced resolution of ambiguities between recognizable commands.
2. Description of the Prior Art
Many electronic devices require input from a user in order to convey to the device particular information required to determine or perform a desired function or, in a trivially simple case, when a desired function is to be performed as would be indicated by, for example, activation of an on/off switch. When multiple different inputs are possible, a keyboard comprising an array of two or more switches has been the input device of choice in recent years.
However, keyboards of any type have inherent disadvantages. Most evidently, keyboards include a plurality of distributed actuable areas, each generally including moving parts subject to wear and damage and which must be sized to be actuated by a portion of the body unless a stylus or other separate mechanical expedient is employed. Accordingly, in many types of devices, such as input panels for security systems and electronic calculators, the size of the device is often determined by the dimensions of the keypad rather than the electronic contents of the housing. Additionally, numerous keystrokes may be required (e.g. to specify an operation, enter a security code, personal identification number (PIN), etc.) which slows operation and increases the possibility that erroneous actuation may occur.
Perhaps more importantly, use of a keyboard inherently requires knowledge of particular keystrokes or combinations thereof which are associated with information (e.g. instructions or data) which must be input. For example, a combination of numbers for actuation of a lock for secured areas of a building or a vehicle requires the authorized user to remember the number sequence as well as correctly actuating corresponding switches in sequence to control initiation of a desired function. Therefore, use of a keyboard or other manually manipulated input structure requires action which is not optimally natural or expeditious for the user. Further, for security systems in particular, the security resides in the limitation of knowledge of a keystroke sequence and not in the security system itself since a security system relying upon correct keypad entry cannot identify the individual person who is currently actuating the keys.
In an effort to provide a more naturally usable, convenient and rapid interface and to increase the capabilities thereof, numerous approaches to voice or sound detection and recognition systems have been proposed and implemented with some degree of success. However, some aspects of known voice actuated systems do not significantly improve over keyboards in regard to transparency to a user or continued security once access is initially granted.
For example, a simple voice actuated system which relies on template matching of the acoustical content of an utterance to allow access to a system (e.g. a file or command set existing on a data processing system) generally requires a particular word or phrase to be input for each enrolled (e.g. authorized) user. (Thus, such a system relying on a particular utterance is referred to as being text dependent or text prompted when the text of an utterance is prompted by the recognizer.) Assuming a correct match is found among enrolled users, no further check of the authorization of the user can be made without requiring the particular recognizable utterance to be repeated during an interruption of usage of the system. Thus, without such interruptions of use, such as by periodically prompting the user for a particular utterance, and the consequent inconvenience, once access is granted, subsequent unauthorized use may not be prevented as might occur, for example, if an authorized user were to momentarily leave a terminal without logging off. It should also be recognized that the requirement for the utterance to be confined to particular text is conceptually identical to the use of a predetermined key sequence such as a personal identification number (PIN) entered on a keypad and, moreover, the utterance cannot be protected from interception by others as readily as a key actuation sequence can be adequately shielded from view. Further, simultaneous authorizations may not be accommodated, much less speech recognition ambiguities resolved in text-dependent systems even though the incidence of such ambiguities would be expected to increase as the population of simultaneously authorized speakers increases. Further, text dependent and text prompted systems can be defeated by mechanical or electrical reproduction of speech. In contrast, continuous checks of speech, making attempts to compromise security difficult together with increasing the difficulty of carrying out any unauthorized use to the point of near impossibility would require text independence.
Indeed, when the speaker recognition is performed in parallel with another process, the command or answers produced by the user must be consistent with the procedure. For example, if the speaker recognition is performed while the user interacts with an operator of a call center, as will be later described, the ability to maintain a coherent discussion is a natural way to check that the voice is not the product of reproduction or synthesis. With automatic systems, answers to questions and relevance of the command fulfill the same role. In a more elaborate case, the system could ask questions personal to the user at some random moment in order to check the answer and thus authenticate the process. On the other hand, when the speaker recognition is text independent, all of the interaction is monitored to verify the user while authenticating the answers.
In summary, dependency of speech recognition on specific text is a limitation on the level of security that can be provided as well as the functionality of the entity to which speaker recognition is applied, whether for security purposes or not. More importantly, however, the obtrusiveness attendant upon text dependency effectively prevents speaker recognition from being used in many applications in which it would yield substantial convenience. For example, once a speaker (or even a combination of speakers) is identified, the identification can be used to retrieve a name and personal information such as preferences, history of previous activity such as commands can be retrieved and used to automatically control a device or system, possibly in an adaptive manner.
Such controls could range from automatic seat and steering wheel adjustments in an automobile to establishing communication links and calling applications commonly used by a particular authorized user when the user logs onto the system. Similarly, text independent speaker recognition would be useful in shared voice dialing systems (corporate voice dialing) and voice controlled answering machines in which the identification of an enrolled speaker could be used to retrieve lists of enrolled names to be dialed or from whom messages should be retrieved (particularly in the case where different speakers might use the same name or word to refer to different people) and other shared “discourse management”arrangements such as voice controlled e-mail, calendars, address books, and the like.
Text independence would also allow the speaker recognition function to be carried out in parallel with other functions in a manner transparent to a user and for a plurality of concurrent users at differing access levels without requiring interruption for new commands or identification of a user whenever a new user is encountered. Thus, access control or personalization can be carried out on a substantially continuous basis without any user detectable interference with other operations or s
International Business Machines - Corporation
Otterstedt Paul J.
{haeck over (S)}mits T{overscore (a)}livaldis Ivars
LandOfFree
Text independent speaker recognition with simultaneous... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Text independent speaker recognition with simultaneous..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Text independent speaker recognition with simultaneous... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2926636