Research And Application Of Market Forecasting Based On Expression Recognition And Classification

With the development of economy, there is an increasing number of purchasing behaviors among people. How to make an effective market forecasting has become consumer companies'focus of attention. The data people get by traditional ways of questionnaire and interview, which are applied to predict market, are often unreliable. Based on the method of facial expression recognition and classification, people can efficiently avoid that and thus get accurate results while making a market forecasting.With facial expression images as the objects of research, the paper attempts to study the face detection and processing, feature extraction, and recognition and classification. Then through statistic analysis of its classification results, the paper helps forecast the market.(1) Face detection and processing of expression images. The face detection and eye location by the use of AdaBoost algorithms and the processing of geometric dimensioning normalization, pixel value standardization, and denoising and enhancement etc. on expression image, ensure that training samples match and contrast with test samples under the same conditions, which effectively improves the recognition rate of expression images.(2) Feature extraction of expression images. Extract the features from an expression image by the use of Gabor filter and choose one from numerous features, which can best reflect its feature contained in the image. Then take amplitude as its eigenvalue so that the efficiency of classifiers'design will be improved.(3) Recognition and classification of expression images. By design of four two-against-two SVM classifiers, six basic facial expressions can be recognized and classified. With the first reasonable designed classifier, error accumulation that results from two-against-two SVM classification can become evitable. Thus, even though there are some errors in the rest three classifiers, people still can forecast the market.(4) Market forecasting based on the results of expression images'classification. By reclassifying the results of six kinds of basic facial expressions and putting different weight on each expression, it will be easily to compute consumer satisfaction to products. And then examine and analyze the validity and reliability of its results so that the goals of market forecasting can be achieved.(5) The design and implementation of a market forecasting system with face detection, eye location, geometric processing, gray equilibrium, feature extraction, identification and classification, and market forecasting as its main functions.