Extraction of the Value Representing Human Feature for the Development of the Sensor with the Function of Privacy-Protection and the Construction of Discrimination-Circuit

In a rest room, the camera cannot be used from the viewpoint of privacy. In such a place, it is important to get the human’s behavior and state without his discrimination. In this study, the sensor that converts the twodimensional image into one-dimensional distribution is proposed. From one-dimensional distribution, the human’s behavior and state can be extracted. As the algorithm, the learning vector quantization (LVQ) is adopted. This is because that the program size is not large, so it is possible to be put in a microcomputer. However, it takes much time to find the optimum parameters of LVQ. For improving it, the S-System was adopted to get the optimum parameters. As a result, the model using LVQ which is called "discrimination circuit" is constructed. By the use of S-System, the apriori knowledge is not needed to get the best parameters.