Real-time Recognition of Piecewise-regular ObjectsAndrey Savchenko
Department of Business Informatics and Applied Mathematics, National Research University High School of Economics, Nizhni Novgorod, Russia
In our presentation we focus on the problem of pattern recognition of composite (piecewise-regular) objects (e.g., speech signals, complex images, etc.). The recognition methods are classified in dependence on the number of available models and the count of classes in the database. We present the mathematical model of composite object as a sequence of independent segments. Each segment is represented as a group (simple random sample) of independent identically distributed feature vectors. Based on this model and statistical approach we reduce the task to a problem of composite hypothesis of testing for segments homogeneity. Several nearest-neighbor criteria are implemented, for some of them the well-known special cases (e.g., the Kullback-Leibler minimum information discrimination principle, the probabilistic neural network) are highlighted. It is proposed to deal with the known issue of insufficient computing efficiency for real-time applications if the number of models per each class is not enough and the number of classes is large (thousands of alternatives) by usage of our directed enumeration method. It is experimentally shown that proposed approach allows to improve the accuracy and decrease the recognition speed in comparison with contemporary approximate nearest neighbor methods.
In our presentation we focus on the problem of pattern recognition of composite (piecewise-regular) objects (e.g., speech signals, complex images, etc.). The recognition methods are classified in dependence on the number of available models and the count of classes in the database. We present the mathematical model of composite object as a sequence of independent segments. Each segment is represented as a group (simple random sample) of independent identically distributed feature vectors. Based on this model and statistical approach we reduce the task to a problem of composite hypothesis of testing for segments homogeneity. Several nearest-neighbor criteria are implemented, for some of them the well-known special cases (e.g., the Kullback-Leibler minimum information discrimination principle, the probabilistic neural network) are highlighted. It is proposed to deal with the known issue of insufficient computing efficiency for real-time applications if the number of models per each class is not enough and the number of classes is large (thousands of alternatives) by usage of our directed enumeration method. It is experimentally shown that proposed approach allows to improve the accuracy and decrease the recognition speed in comparison with contemporary approximate nearest neighbor methods.
Have you spotted a typo?
Highlight it, click Ctrl+Enter and send us a message. Thank you for your help!
To be used only for spelling or punctuation mistakes.