Research on Feature Extraction and Classification of Tongue Shape and Tooth-Marked Tongue in TCM Tongue Diagnosis
|School||Harbin Institute of Technology|
|Course||Computer Science and Technology|
|Keywords||Tongue Diagnosis Feature Extraction Fourier Descriptor BP Neural Network Particle Swarm Optimization|
Tongue Diagnosis is an important part of Traditional Chinese Medicine (TCM). It obtains information through observing the tongues’ characteristics of patients to get their state of healthiness, and then works out diagnosis and state of an illness. It is one of the most valuable and widely used diagnostic methods in Traditional Chinese Medicine. Meanwhile, as the development of the modern science and technology, it is important and valuable for the development and application of Tongue Diagnosis to combine the computer technology such as digital image processing and pattern recognition and Tongue Diagnosis in order to avoid the defect of inquantitativeness and subjectiveness of Tongue Diagnosis. The purpose of this dissertation is contributing to the computerization of TCM tongue diagnosis by introducing feature extraction method to TCM tongue diagnosis research. The main contents of this thesis are following:(1) This dissertation introduces Fourier descriptor, BP neural network and Particle Swarm Optimization. This dissertation explains the method of object shape identification by using Fourier descriptor, then introduces the learning process of BP neural network, at last analyzes the advantage of Particle Swarm Optimization.(2) This dissertation gets tongue marginal points by Snakes arithmetic, then presents a novel arithmetic of marginal point equalization and does it in tongue shape. The traditional BP neural network is prone to fall to local optimality, to a certain extent, affecting the performance of the tongue shape identification system. Therefore Particle Swarm Optimization is introduced to traditional BP neural network learning algorithm to improve it. Compared with the traditional BP neural network, the improved BP neural network learning algorithm can not only effectively converge to the globe optimum, but also gives better generative performance.(3) This dissertation introduces the cause of Tooth-marked tongue and its definition. A property-based approach is presented for shadow removal of tongue image. Then this dissertation extracts the edge feature, color feature of tongue side and the color feature of tongue surface. At last, this dissertation combines these three features with the feature of tongue shape, constructs linear classifier to identify the tooth-marked tongue.