Masafumi Hagiwara

A feeling estimation system using a simple electroencephalograph

This paper proposes a system for feeling estimation. It uses a simple electroencephalograph as an interface. If quantitative feeling presumption with information from a living body is made possible, it can be considered that a computer can appropriately deal with humans natural sensibility. Input data of proposed system is electroencephalogram (EEG) data of several feelings acquired using a simple electroencephalograph. Output of the system is one of some words associated with certain feelings states. One feature of proposed system is performing classification using the neural network. The neural networks have generalization ability and can deal with data including noise. Another feature is a low burden for the user by using a simple electroencephalograph. The proposed system is applicable to the quantitative evaluation replaced with a questionnaire. It is also applicable to handling the non-verbal information that is missing at communication.

Fuzzy inference neural network

Abstract A new model for the design of Fuzzy Inference Neural Network (FINN) is proposed in this paper. It can automatically partition an input-output pattern space and can extract fuzzy if-then rules from numerical data. The proposed FINN is a two-layer network which utilizes Kohonens algorithm. There are three learning phases: self-organizing learning phase, rule-extracting phase, and supervised learning phase. The FINN has the following distinctive features: (1) the membership functions of the premise part are constructed in the connection between the input layer and the rule layer; (2) it has an ability to select a suitable number of rules adaptively; and (3) it can extract more refined fuzzy if-then rules. We apply the proposed FINN to two illustrative examples, fuzzy control of an unmanned vehicle, and the prediction of the trend of stock prices. Computer simulation results indicate the effectiveness of the FINN.

Adaptive fuzzy inference neural network

An adaptive fuzzy inference neural network (AFINN) is proposed in this paper. It has self-construction ability, parameter estimation ability and rule extraction ability. The structure of AFINN is formed by the following four phases: (1) initial rule creation, (2) selection of important input elements, (3) identification of the network structure and (4) parameter estimation using LMS (least-mean square) algorithm. When the number of input dimension is large, the conventional fuzzy systems often cannot handle the task correctly because the degree of each rule becomes too small. AFINN solves such a problem by modification of the learning and inference algorithm.

Image query by impression words-the IQI system

A new image query system by impression words named the IQI system has been proposed. The proposed IQI system has been implemented on a Sun SS-5 workstation. Since human sensitivity and various images are flexibly connected using a neural network in the IQI system, it can automatically estimate impression words from various kinds of images. As a result, the IQI system can reflect a humans kansei (impression and sensitivity) effectively and treat a wider variety of images. We show some experimental results. The IQI system can estimate correct impression words and successfully retrieve correct images.

Separation of superimposed pattern and many-to-many associations by chaotic neural networks

We propose a chaotic associative memory (CAM). It has two distinctive features: 1) it can recall correct stored patterns from superimposed input; and 2) it can deal with many-to-many associations. As for the first feature, when a stored pattern is given to the conventional chaotic neural network as an external input, the input pattern is continuously searched. The proposed model makes use of the above property to separate the superimposed patterns. As for the second feature, most of the conventional associative memories cannot deal with many-to-many associations due to the superimposed pattern caused by the stored common data. However, since the proposed model can separate the superimposed pattern, it can deal with many-to-many associations. A series of computer simulations shows the effectiveness of the proposed model.

Successive learning in chaotic neural network

In this paper, we propose a successive learning method in a chaotic neural network using a continuous pattern input. It can distinguish an unknown pattern from the stored known patterns and learn the unknown pattern successively. In the proposed model, it makes use of the difference in the response to the input pattern in order to distinguish an unknown pattern from the stored known patterns. When an input pattern is regarded as an unknown pattern, the pattern is memorized. Furthermore, it can estimate and learn a correct pattern from a noisy unknown pattern or an incomplete unknown pattern by considering the temporal summation of the continuous pattern input. In addition, similarity to the physiological facts in the olfactory bulb of a rabbit found by Freeman (1991) is observed in the behavior of the proposed model. A series of computer simulations shows the effectiveness of the proposed model.

Scenery image recognition and interpretation using fuzzy inference neural networks

Abstract In this paper, we propose a new image recognition and interpretation system. The proposed system is composed of three processes: (1) regional segmentation process; (2) image recognition process; and (3) image interpretation process. As a pre-processing in the regional segmentation process, an input image is divided into some proper regions using techniques based on K -means algorithm. In both the image recognition and the interpretation processes, fuzzy inference neural networks (FINNs) working in parallel are employed to achieve a high level of recognition and interpretation. Scenery images are used and it is confirmed that the system has an average of 71.9% accuracy rate in the recognition process and good results in the interpretation process without heuristic knowledge. In addition, it is also confirmed that the proposed system has an ability to extract proper rules for the image recognition and interpretation.

A combined multi-winner multidirectional associative memory

Abstract We propose a new multidirectional associative memory (MAM) termed combined multi-winner MAM (CMW–MAM) using a distributed representation paradigm in this paper. The proposed CMW–MAM stores and recalls information hierarchically by the distributed representations which are automatically generated in the networks. The proposed CMW–MAM can handle analog information and it can store and recall composite many-to-many memory structures such as episodes-to-episodes associations. The association properties for the storage of composite many-to-many memory structures such as associations from some episodes to other episodes are studied. The basic properties for many-to-many association such as storage capacity, noise performance and robustness are also investigated in detail.

Episodic associative memory

Episodic associative memory (EAM) is introduced and simulated. It uses quick learning for bidirectional associative memory (QLBAM) and pseudo-noise (PN) sequences. The features of the proposed EAM are: it can memorize and recall episodic associations; it can store plural episodes; it has high memory capacity.<<ETX>>

Large scale on-line handwritten Chinese character recognition using improved syntactic pattern recognition

We propose an original method for the recognition of online handwritten Chinese characters using an improved syntactic pattern recognition. Syntactic pattern recognition is a method that converts a pattern into a string of symbols using a finite set of features and then analyzes them structurally using grammars. It is effective for such patterns as structurally constructed Chinese characters. We use Kohonens self-organizing feature map for feature extraction, to get optimal sets of prototypical waveforms of peaks from sample data automatically. The strings of symbols are converted into matrices which express features of the successors, and are analyzed by simple calculations between matrices. Moreover in order to symbolize and analyze efficiently and accurately in a large scale, we employ a hierarchical approach for the proposed method. Using free writing characters, we obtained a 99.49% recognition rate for training patterns and 94.34% for test patterns.