T 0929/15 - Identifying a user issuing a voice request

Key points

• Claim 1 is directed to a " Computer network for natural language-based control of a digital home network". 
• The relevant distinguishing feature is: "wherein the voice identification component (220) is operable to identify users issuing user requests by: - processing incoming voice samples of each user request; - extracting features from the incoming voice samples; and - matching the extracted features against voice prints of users stored in a database" 
• The Boar: " The distinguishing feature that is worth further consideration is that the processing of a natural language-based voice request, besides producing a list of tags, also results in the identification of the user that provided the request. In particular, the user is identified by processing incoming voice samples of each request, extracting features therefrom and matching the extracted features against stored voice prints of users." 
• "The solution suggested in claim 1 is to biometrically identify the user who issued the voice request, based on features of the voice sample." 
• The Board finds the feature to be inventive: " the appellant convincingly argued that even though [the closest prior art document] D4 suggests [] that user authentication be required in order to access a companion - an authentication which might also take the form of an audio signature or a biometric signature - D4 consistently teaches (e.g. paragraph [0035]) that in such cases user identification is a precondition for accepting any request from the user. This teaches away from identifying the user while the received request is being processed.”
• The other documents do not teach the feature either.

EPO T 0929/15 -  link

2. Inventive step (Article 56 EPC)

2.1 Document D4 represents the closest prior art to the subject-matter of claim 1.

2.2 Claim 1 comprises a number of features which are not disclosed in D4. The distinguishing feature that is worth further consideration is that the processing of a natural language-based voice request, besides producing a list of tags, also results in the identification of the user that provided the request. In particular, the user is identified by processing incoming voice samples of each request, extracting features therefrom and matching the extracted features against stored voice prints of users.

2.3 This distinguishing feature has the technical effect that each voice request is associated with the identity of the user who issued it, without the need for the user to explicitly input their identity.

2.4 The objective technical problem solved by this feature can thus be regarded as how to reliably and conveniently map a voice request to the user who issued the request.

2.5 The solution suggested in claim 1 is to biometrically identify the user who issued the voice request, based on features of the voice sample.

2.6 D4 suggests in several paragraphs, notably [0089] and [0092], that a group of users can each have a respective companion, or make use of a central companion running multiple processes for each user. However, it is silent on the complications that may arise in these scenarios, in particular how the companion(s) is/are to map a received voice request to a particular user. The appellant convincingly argued that even though D4 suggests, notably in paragraph [0094], that user authentication be required in order to access a companion - an authentication which might also take the form of an audio signature or a biometric signature - D4 consistently teaches (e.g. paragraph [0035]) that in such cases user identification is a precondition for accepting any request from the user. This teaches away from identifying the user while the received request is being processed.

2.7 Therefore, the claimed solution would not be obvious to the skilled person based on D4 alone. Nor would the skilled person be guided by the remaining documents on file to arrive at the solution in claim 1, since these documents are either entirely silent on providing a voice user interface or do not suggest user identification based on voice.

2.8 Consequently, based on available prior art, the subject-matter of claim 1 of the new main request involves an inventive step (Article 56 EPC).