Data processing: speech signal processing – linguistics – language – Speech signal processing – Application
Reexamination Certificate
2002-08-12
2004-08-17
{grave over (S)}mits, T{overscore (a)}livaldis Ivars (Department: 2655)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Application
C704S270000
Reexamination Certificate
active
06778964
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a configuration with an electrically operated appliance with a voice input unit and with a voice processing unit that derives necessary control signals from spoken input instructions for controlling functions of the appliance, and a voice input method with correction of interference signals.
International Publication WO 98/55992, corresponding to U.S. Pat. No. 6,230,137 to Has et al., discloses a configuration and a method in which a household appliance is described that can be controlled by voice signals. To avoid malfunctions, for example, due to voice signals that do not originate from an operator but rather, for example, from a radio, the publication describes configuring a control program such that successive voice signals lead to the formation of a control instruction only if the successive voice signals are input within a predefinable time.
Furthermore, a product with the designation “SICARE pilot” (from evosoft Software-Vertrieb [sales] GmbH, D-90411 Nuremberg/Germany) is already in the REHA field and includes a mobile appliance that converts spoken instructions into signals. The signals are output by an infrared transmitter, for example, for controlling television appliances, video recorders, hi-fi equipment, telephone and lighting, by a radio transmitter, for example, for controlling house doors, wheelchairs, personal paging systems, and alarm systems or by wire-bound interfaces, for example, for controlling electrically driven wheelchairs and beds that can be adjusted by motor. The prior art product, which is intended for disabled people, permits a maximum of 64 instructions to be stored. The mobile appliance transmits the instruction for the desired action, for example, “television on,” to the respective appliance automatically through the previously selected interface. Each individual instruction can address a plurality of appliances simultaneously. The prior art product is trained to the individual voice pattern of a person and, thus, carries out speech recognition on a speaker-dependent basis. In an ideal case, other persons are not able to form signals that control the appliance from spoken instructions.
Such voice recognition is implemented in the prior art product such that predefined words (for example “house,” “light,” “four,” or “minus”) can be spoken in accordance with a menu tree with a plurality of branches, it being possible to speak the individual words at any desired intervals. If the voice input is prematurely aborted as a result of an indisposition of the disabled person, individual words, for example, from a radio or television, which are provided in the branch of the menu tree after the point at which the voice input was aborted may lead to the formation of signals that control one or more appliances.
Furthermore, German Published, Non-Prosecuted Patent Application DE 196 37 352 A1 discloses a method for automatic voice recognition, reference utterances being firstly recorded for different words and then sequences of chronologically successive reference feature vectors being formed from them, which vectors are each mapped onto a uniform chronological interval. A single model of the reference utterances, whose components are composed of compensation functions, is formed for each word from the sequences of model vectors that are produced here for the reference utterances. A word that is to be recognized is processed to form a sequence of feature vectors that are each mapped onto the same chronological interval. The resulting sequence of mapping vectors is compared with the stored models in comparison steps.
Furthermore, the manual “Sprachsynthese und Spracherkennung [speech synthesis and speech recognition]”, J. N. Holmes, R. Oldenbourg-Verlag, Munich, Vienna 1991, generally discloses that some information items relating to the phonetic content of the speech signal can be derived from the chronological profile of the speech signal. However, experience has shown that the time signal is not suitable for representing the properties of spoken language that are most important for general voice quality and perception of phonetic details. In view of the significance of the resonances and their chronological changes during verbal communication, a method for representing these features is necessary. The short-term spectrum of the signal that corresponds to the absolute value of the Fourier transformation of the time signal after it has been multiplied by a time window function with an appropriate duration may, of course, not contain any information that is not also contained in the original signal. However, it is also significantly more suitable for representation of the resonances and, in particular, for the computer-supported signal processing.
SUMMARY OF THE INVENTION
It is accordingly an object of the invention to provide an electrical appliance with voice input unit and voice input method that overcome the hereinafore-mentioned disadvantages of the heretofore-known devices and methods of this general type and that improves speech recognition despite the presence of interference noises.
With the foregoing and other objects in view, there is provided, in accordance with the invention, a configuration having an electrically operated household appliance, a voice input unit, and a voice processing unit deriving control signals from spoken input instructions for controlling functions of the household appliance, including an operating status detection unit to be connected to the household appliance and the voice processing unit, the operating status detection unit adapted to detect an operating status of at least one of the household appliance and other noise sources, to signal the operating status of the at least one of the household appliance and other noise sources to the voice processing unit, and to control the voice processing unit to perform an interference noise correction only if at least one of the noise sources is on.
With the objects of the invention in view, there is also provided a method for inputting voice signals with correction of interference signals, including the steps of interrogating an operating status of at least one noise source interfering with voice input during voice input for controlling a household appliance and performing an interference noise correction with a voice processing unit only if a noise source is on.
The invention provides an operating status detection unit that detects the operating status of the household appliance or of other noise sources and signals it to the voice processing unit, and that the voice processing unit performs an interference noise correction only if a noise source is switched on. The method according to the invention for voice input with correction of interference signals is characterized in that the operating status of at least one noise source that interferes with the voice input is interrogated during the voice input for controlling an appliance, and the voice processing unit performs an interference noise correction only if a noise source is switched on. According to the invention, therefore, when a voice signal for detection is submitted to the voice processing unit, an interference noise correction is not attempted in every case. Such a process leads to an improvement in the voice recognition in all the cases in which the voice signal is not at all subjected to interference noises. This is because the attempt to remove a nonexistent interference noise from the voice signal, or to include it in a correcting fashion, reduces the quality of the voice signal. This causes the recognition rate of the voice processing unit or voice recognition unit in the prior art to drop.
In accordance with another feature of the invention, the voice processing unit includes the interference signal of the noise source in a correcting fashion in the processing of the spoken input instructions, as a function of the respective operating status of the noise source.
For such a purpose, in accordance with a fur
Geiger Dieter
Haberlander Rainer
Has Uwe
BSH Bosch und Siemens Hausgerate GmbH
Greenberg Laurence A.
Jackson Jakieda
Mayback Gregory L.
Stemer Werner H.
LandOfFree
Electrical appliance voice input unit and method with... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Electrical appliance voice input unit and method with..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Electrical appliance voice input unit and method with... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3333088