The Applications Of Artificial Intelligence Technique In The Signal Characteristics’ Analysis

• Main Authors: Yang, Jenpin, 楊仁賓
• Other Authors: Hwang, Reychue
• Format: Others
• Language: en_US
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/85116304305507316695

博士 === 義守大學 === 電機工程學系 === 100 === This thesis presents the artificial intelligence (AI) technique in signal characteristics’ analysis and its applications. Four different research topics were studied. First research is the indoor positioning system (IPS) design. In this part, five IPS techniques including triangulation method, feed-forward neural network (NN), modified probabilistic neural network, particle swarm optimizer, and fuzzy logic theory were studied and proposed. This research aims to design an IPS with high stability, high accuracy and high reliability. Through the experiments, the advantages and capabilities of different AI techniques in IPS could be clearly compared and discriminated. The second research is the estimation of ammonia concentration of shear horizontal surface acoustic wave sensor (SH-SAW). In this part, the data sensed by SH-SAW sensors were implemented by using different neural network models. A reliable and superior neural network SAW identifier is expected to be created for effectively overcoming the interference of humidity in ammonia detection. The third research is the estimations of rolled steel bar’s mechanical properties. In this part, the estimator of mechanical properties of rolled steel bar by using quantum neural network (QNN) was developed. Based on this QNN estimator, the control parameters of manufacturing process of steel bar can be set precisely in advance. Thus, such an AI estimator can not only improve the quality of steel bar, but also reduce the cost of bar’s production. The fourth research is the prediction of electric contract capacity. In this part, an AI predictor of electric contract capacity for community antenna television system (CATV) based on QNN technique was developed. Based on the powerful learning capability of NN, the nonlinear and complex relationship between power demand and its possible influencing factors could be automatically obtained. Thus, such a well-trained neural model could be employed into the power demand prediction accurately. This thesis presents the different AI techniques in the signals with different characteristics. Then, the particular application can be investigated based on individual signal characteristic. The experimental results show that the AI techniques employed in different signal analyses and applications are feasible and promising. From the study results shown, the AI techniques do have the great improvement in the application’s efficiency and accuracy in comparison with the traditional approaches.