Apparatus, system and method for voice dialogue activation and/or conduct

摘要

An apparatus, a system and a method for voice dialogue activation and/or conduct. The apparatus for voice dialogue activation and/or conduct has a voice recognition unit, a speaker recognition unit and a decision-maker unit. The decision-maker unit is designed to activate a result action on the basis of results from the voice and speaker recognition units.

主权项

1. An apparatus for at least one of voice dialogue activation and voice dialogue conduct, for use in a vehicle, comprising:
at least one input for a voice signal; a voice recognition unit configured to establish one or more command words contained in the voice signal; a speaker recognition unit configured to determine a current speaker using the voice signal and at least one stored speaker profile; a decision-maker unit comprising:
a voice recognition unit connection coupled to an output of the voice recognition unit configured to perform a result action based on the one or more command words, anda speaker recognition unit connection coupled to the speaker recognition unit,the decision-maker unit being configured such that the activation of the result action is dependent, at least in the case of at least one command word, on whether the at least one command word has been identified as coming from a speaker associated with a speaker profile; and an echo cancellation unit that receives a multichannel voice signal and, on the basis of transit time differences among components of the multichannel signal with respect to the at least one input, removes all components from non-authorized speakers, wherein: the speaker recognition unit is configured to identify the current speaker by extracting speaker features from the voice signal and comparing the speaker features with stored speaker-dependent features, and comprises a further unit configured for speaker adaptation to continually ascertain refined speaker-dependent features and store the refined speaker-dependent features in the stored speaker profiles, and the speaker recognition unit is configured to, in the case that a plurality of speakers are speaking simultaneously, attribute the voice signal to no speaker.