الفهرس 
Chapter1: IntroductionOnly 14 pages are availabe for public view
 |  | from | 213 |  |  |
| | | from | 213 | | |
Abstract
There is an increasing number of children in primary and pre-schools with overcrowding in the areas where educational institutions exist. Children may be exposed to many risks such as accidents, loss, and fainting caused by any chronic diseases. Therefore, there is a huge need to recognize the pupils to know their information which enables us to help them. This study presents a multi-agent system for the mobile device to recognize pupils. The system consists of three agents : image recognition agent, voice recognition agent, and knowledge base agent. Image recognition agent captures the pupil image with a mobile camera then detects his face and recognizes him. Voice recognition agent records the voice using a mobile microphone, extracts features, and decides the child’s identity. The knowledge base agent receives information from a user, compares it with facts, and applies rules to decide the pupil’s identity. Every mentioned agent takes some actions after the decision like show personal data of pupils on screen or suggesting first aids for some cases. The system used algorithms such as eigenfaces, Gray-Level Co-occurrence Matrix (GLCM) for image processing, Mel-Frequency Cepstral Coefficients (MFCC) and Linear Predictive Coding (LPC) for sound processing. The system was implemented on a smartphone device. System was applied to datasets of faces and voices. The images of dataset were captured for primary stage pupils in different conditions like anger, normal, laugh and sadness. The voices of pupils were recorded in the same conditions also. The confusion matrix was used to evaluate the proposed system accuracy. The results showed that the accuracy of the proposed system of the eigenfaces method was 94% and the GLCM method was 95 %. So, the recognition rate of the proposed system with the two algorithms is very high and efficient. The accuracy of the MFCC and LPC methods was 95%, so there were no significant differences between the two methods. The recognition rate of the proposed system by the voice agent is precise by the two methods. The accuracy for a knowledge-based agent was 81%, this agent gave an acceptable ratio.