Action evaluation apparatus and method

• US 20060293900A1
• Filed: 06/21/2006
• Published: 12/28/2006
• Est. Priority Date: 06/22/2005
• Status: Active Grant

• Current Assignee: Sony Corporation (Sony Group Corp.)
• Original Assignee: Sony Corporation (Sony Group Corp.)
• Inventors: Sawada, Tsutomu
• Granted Patent: US 7,523,084 B2
• Time in Patent Office: Days
• US Class Current: 704/275
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

The present invention provides an action evaluation apparatus that evaluates a user'"'"'s action corresponding to an action desired by a system, and determines ongoing of a story using the evaluation result. There is provided an action evaluation apparatus for evaluating a user'"'"'s action, including a recognizer for recognizing the user'"'"'s action through a plurality of modals, and an action evaluator for evaluating the user'"'"'s action based on the recognition result of the respective modals obtained by the recognizer.

Claims

1. An action evaluation apparatus for evaluating a user'"'"'s action, comprising:

a recognition means for recognizing the user'"'"'s action through a plurality of modals; and

an action evaluation means for evaluating the user'"'"'s action based on the recognition result of the respective modals obtained by the recognition means.

2. The action evaluation apparatus as set forth in claim 1, wherein the action evaluation means evaluates the user'"'"'s action using instructor data for the respective modals.

3. The action evaluation apparatus as set forth in claim 2, wherein

the recognition result data and the instructor data are expressed in the form of a vector, and

the action evaluation means evaluates the user'"'"'s action based on an inter-vector distance between the recognition result data and the instructor data.

4. The action evaluation apparatus as set forth in claim 1, wherein

the action evaluation means calculates, for the user'"'"'s action, action evaluation values for the respective modals and an overall evaluation value through all the modals, and

further comprises a notification means for, in case the overall evaluation value is less than a predetermined threshold value, notifying the user of action correction points for modals whose action evaluation value is less than a predetermined threshold value.

5. The action evaluation apparatus as set forth in claim 1, wherein the action evaluation means calculates a matching probability of the user'"'"'s action and a target action using the Bayesian Network, and evaluates the user'"'"'s action based on the matching probability.

6. The action evaluation apparatus as set forth in claim 1, wherein the recognition means uses, as modals, at least two items of the face position and direction, visual line direction, face expression, movement of hands and legs, shape of mouth, shape of eyes, shape of eyebrows, posture, contents of speech, and rhythm.

7. An action evaluation method for evaluating a user'"'"'s action, comprising the steps of:

recognizing the user'"'"'s action through a plurality of modals; and

evaluating the user'"'"'s action based on the recognition result of the respective modals obtained in the recognition step.

8. An action evaluation apparatus for evaluating a user'"'"'s action, comprising:

a recognizer for recognizing the user'"'"'s action through a plurality of modals; and

an action evaluator for evaluating the user'"'"'s action based on the recognition result of the respective modals obtained by the recognizer.