JPH1049707A - Human body reproducing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform subjective and correct human body reproduction with high reproduction quality by detecting the expression of a human body in real time, preserving the result, also preserving data for reproducing the human body in real time, learning them so as to relate them and reproducing the expression and operation of the human body by using a learned result. SOLUTION: A real time expression detected by a real time expression detection circuit l is stored in a real time expression detection storage circuit 3. In the meantime, real time information reproduced by a real time expression reproducing device 2 is stored in a real time expression reproduction storage circuit 4. Based on the detected real time information stored in the real time expression detection storage circuit 3 and the reproduced real time expression stored in the real time expression reproduction storage circuit 4, a learning circuit 5 learns the relation of both. In this case, for the expression reproduced by a computer graphic, the relation with the actual expression is learned. Then, after learning is ended, reproduction is performed based on learned data.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a person reproducing method, and more particularly to a state reproducing method for reproducing a state when an object moves arbitrarily. The present invention relates to a method for reproducing a person that can be reproduced on a personal computer.

[0002]

2. Description of the Related Art FIG. 3 is a block diagram showing a conventional expression reproducing method, and FIG. 4 is a diagram showing an example of reproducing an expression.

[0003] A non-contact input device such as a CCD camera captures a facial expression of a person as a three-dimensional model.
The facial expression is detected by the real-time facial expression detection circuit 1, and based on the detection output, the facial expression of the human face is reproduced by the real-time facial expression reproducing device 2 and displayed on a display or reproduced by a robot.

[0004]

However, in the conventional expression reproducing method, even if the three-dimensional expression is faithfully performed, the human cannot necessarily reproduce the expression with the fidelity such that both are recognized as the same expression. In addition, there are individual differences in the facial expressions that each person expresses, and the facial expression intended by the individual is not always a facial expression that can be understood by other people.

For example, in the conventional facial expression reproducing method shown in FIG. 3, a face of a person detected in real time as shown in FIG. When graphically displayed, as shown in FIG. 4B, there may be a case where the facial expressions are reproduced in such a way that they are received with different facial expressions even though they are the same as the actual facial expressions. In addition, even if the expression of a person who becomes a model laughs, the expression may not be reproduced as if the expression was laughed by other people due to the problem of individual differences.

[0006] Therefore, a main object of the present invention is to provide a person reproducing method capable of realizing subjectively correct person reproduction with higher reproduction quality.

[0007]

The invention according to claim 1 is
This is a person reproduction method that detects and reproduces facial expressions and movements of a person, and saves the results of detecting facial expressions and movements of a person in a non-contact form in real time, and reproduces the person in real time. Is stored in a time-series manner, learning is performed so as to associate the stored data with each other, and facial expressions and actions of a person are reproduced using the learned result.

[0008]

1 and 2 are block diagrams showing an embodiment of the present invention. FIG. 1 shows a learning state.
FIG. 2 shows a reproduction time.

In FIG. 1, a real-time facial expression detected by a real-time facial expression detection circuit 1 is stored in a real-time facial expression detection storage circuit 3.
Is stored. On the other hand, real-time information reproduced by the real-time expression reproduction device 2 is stored in the real-time expression reproduction storage circuit 4. Based on the detected real-time information stored in the real-time facial expression detection storage circuit 3 and the reproduced real-time facial expression stored in the real-time facial expression reproduction storage circuit 4, the learning circuit 5 relates the two. To learn.

Here, the expression reproduced by computer graphics is deformed (deformed, exaggerated) based on art anatomy in advance, and the relation with the actual expression is learned. In this learning, GA (Genetic Algorithm), HMM (Hidden Markov Mode)
l) Learning using neuro, least squares method, etc.

After the learning is completed, as shown in FIG.
The association circuit 6 performs reproduction based on the learned data. As described above, when the facial expression is reproduced using the actual facial expression change and the reproduction result deformed based on the subjectivity in advance, subjectively correct facial expression reproduction is realized. For example, in the case of Kabuki that requires a special facial expression during facial expression changes, it is extremely difficult for ordinary people to actually express the facial expressions of Kabuki actors, but Kabuki actors do. The facial expression is created in advance by computer graphics, and the subject expresses this facial expression. Even if the expression that is expressed is different from the expression of the Kabuki actor, the face of the Kabuki actor can be actually reproduced by learning the relationship between the two. Similarly, even if the expression to be reproduced is not a human such as a kabuki actor, the same result can be obtained by using the expression of an animal such as a dog or a cat.

[0012]

As described above, according to the present invention, the expression and movement of a person are detected in a non-contact manner in real time, the results are stored in a time series, and the person is reproduced in real time. Data is stored in chronological order, and learning is performed so as to associate the stored data with each other, and the expressions and movements of the person are reproduced using the learned result, realizing a higher-quality reproduction of the person. can do.

[Brief description of the drawings]

FIG. 1 is a block diagram at the time of learning according to an embodiment of the present invention.

FIG. 2 is a block diagram at the time of reproduction of an embodiment of the present invention.

FIG. 3 is a diagram showing a conventional expression reproducing method.

FIG. 4 is a diagram showing a conventional facial expression reproduction example.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 real-time facial expression detection circuit 2 real-time facial expression reproducing device 3 real-time facial expression detection storage circuit 4 real-time facial expression reproduction storage circuit 5 learning circuit 6 association circuit

 ────────────────────────────────────────────────── ─── Continuing on the front page (72) Jun Kurumizawa, Inventor 5 Shiraya, Inaya, Seika-cho, Soraku-cho, Kyoto Prefecture AT / IR Intelligent Motion Imaging Communications Laboratory (72) Inventor Atsushi Ohtani Atsushi Kyoto No. 5, Hiratani, Seiya-cho, Gunma-cho, Gunma ATI Intelligent Imaging Communications Laboratory

Claims (1)

[Claims] 1. A person reproducing method for detecting and reproducing a facial expression or a motion of a person, wherein a result of detecting the facial expression or the motion of the person in a non-contact form in real time is stored in a time series, and Data for reproducing a person in real time is stored in chronological order, learning is performed so as to associate the stored two, and the expression and motion of the person are reproduced using the learned result. The person reproduction method.