Voice dialog device and voice dialog method

• US 10,438,586 B2
• Filed: 05/18/2017
• Issued: 10/08/2019
• Est. Priority Date: 05/31/2016
• Status: Active Grant

• Current Assignee: Toyota Jidosha Kabushiki Kaisha (Toyota Motor Corporation)
• Original Assignee: Toyota Jidosha Kabushiki Kaisha (Toyota Motor Corporation)
• Inventors: Ikeno, Atsushi, Shimada, Muneaki, Hatanaka, Kota, Nishijima, Toshifumi, Kataoka, Fuminori, Tonegawa, Hiromi, Umeyama, Norihide
• Primary Examiner(s): Leland, III, Edwin S
• Application Number:
  US15/598,478
• Publication Number:
  US 20170345424A1
• Time in Patent Office: 873 Days
• Field of Search: 704251
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 15/24 Speech recognition using non-acoustical features
• CPC Class Codes: G10L 2015/227 of the speaker; Human-factor methodology
• CPC Class Codes: G10L 25/63 for estimating an emotional state

Abstract

A voice dialog device, comprises a voice processing unit configured to obtain a voice pronounced by a user and a result of recognizing the voice; a plurality of estimation units configured to estimate emotion of the user by different methods; and a response unit configured to create a response sentence, based on results of estimating the emotion of the user, and provide the response sentence to the user, wherein when a discrepancy exists between the results of estimating the emotion of the user by the plurality of estimation units, the response unit makes an inquiry to the user, and determines which estimation result is to be adopted, based on content of an obtained response.

Claims

1. A voice dialog device, comprising a processor configured to perform steps of:

a voice processing step of obtaining a voice pronounced by a user and a result of recognizing the voice;

a plurality of estimating steps of estimating emotion of the user by different estimation methods; and

a responding step of creating a response sentence, based on results of estimating the emotion of the user, and providing the response sentence to the user, wherein

in the responding step, when a discrepancy exists between the emotions of the user estimated by the plurality of estimation methods respectively, an estimation history that is results of estimation of emotion performed at a previous time is obtained, and the discrepancy is resolved based on the estimation history.

2. The voice dialog device according to claim 1, wherein

the estimation history includes true or false information regarding whether a correct result has been obtained as a result of executing estimation of emotion by the respective estimation methods at the previous time, and

in the responding step, weighting for the respective estimation methods is performed based on the true or false information.

3. A voice dialog method executed by a voice dialog device, the method comprising:

a voice processing step of obtaining a voice pronounced by a user and a result of recognizing the voice;

a plurality of estimating steps of estimating emotion of the user by different estimation methods; and

a responding step of creating a response sentence, based on results of estimating the emotion of the user, and providing the response sentence to the user, wherein

in the responding step, when a discrepancy exists between the emotions of the user estimated by the plurality of estimation methods respectively, an estimation history that is results of estimation of emotion performed at a previous time is obtained, and the discrepancy is resolved based on the estimation history.

4. The voice dialog method according to claim 3, wherein

the estimation history includes true or false information regarding whether a correct result has been obtained as a result of executing estimation of emotion by the respective estimation methods at the previous time, and

in the responding step, weighting for the respective estimation methods are performed based on the true or false information.

5. A non-transitory computer readable storing medium recording a computer program for causing a computer to perform a method comprising:

a voice processing step of obtaining a voice pronounced by a user and a result of recognizing the voice;

a plurality of estimating steps of estimating emotion of the user by different estimation methods; and

a responding step of creating a response sentence, based on results of estimating the emotion of the user, and providing the response sentence to the user, wherein

in the responding step, when a discrepancy exists between the emotions of the user estimated by the plurality of estimation methods respectively, an estimation history that is results of estimation of emotion performed at a previous time is obtained, and the discrepancy is resolved based on the estimation history.