Abstract
In this paper, we propose an individual discernment method using the speech sound and movement feature of the lip in speech. To compute the similarity obtained from the input data and registration data in the sound and image data, respectively, the proposed method uses DP matching processing. The method takes not only his or her registration data but also other registration data into consideration; it is able to decrease a non-registrant's number of incorrect receipts without increasing a registrant's number of incorrect rejections.
A case study was carried out on condition that the range of weight coefficient for the sound and image data was 0 to 1.0 at every 0.1. When ten subjects (person a to person j) were registrants, the proposed method is possible to distinguish them perfectly from others. Also 99.3% cased has rightly estimated in case of all subjects were divided into registrants and non-registrants. The experimental results indicate that the proposed method is useful.
