/ Hikaru Date / Professor
/ Qiangfu Zhao / Associate Professor
/ Jintae Lee / Assistant Professor
The multimedia devices presently available add sound, still pictures, and moving pictures to traditional computer displays which show only text and diagrams. The addition of these few functions, however, offers remarkable convenience to the user of such multimedia devices. For this reason, it is widely believed that the impact of these devices on human society in the near future will be extremely great.
The properties desirable for multimedia devices can be visualized on a graph with two axes. The horizontal axis represents a device's capacity for depth and variety of communication and the vertical axis represents the capacity for naturalness and high fidelity. Ideally, multimedia devices should be far along both axes if they are to offer the most comfortable and efficient human interfaces. Unfortunately, progress along these dimensions has not gone far enough. Prolonged use of computers still fatigues the body and fails to match the depth and quality of communication that can exist between humans in face-to-face encounters.
For example, human beings can communicate with great ease and variety through the spoken word. Computers can't. In spite of the fact that speech synthesis by computers is almost perfect, speech recognition has not reached the level of practical application that natural speech can achieve. Furthermore speech recognition algorithms are overly sensitive to existing noise. The area of speech recognition is still in the primitive stages and needs to evolve further to make multimedia devices more effective.
Future multimedia systems should be able to accept and recognize natural information. While traditional multimedia systems provide more and more information, future systems should be able to select, and provide useful and important information. One project in this laboratory is to study how to build such kind of intelligent multimedia systems. Our short-term goal is to build a system that can recognize 3D characters and documents. The long-term goal is to construct a system that can recognize nature images and speeches.
Sign language is another means of human communication that computers cannot participate in as easily as those who are forced to use it because of hearing impairment. If a computer could be made to communicate via this medium, far more people could access intellectual services and actively participate in multimedia that have been unable to do so thus far. At present, pioneer research to incorporate this dimension of communication as far as synthesis and recognition technology has just begun in our laboratory.
As for the dimension of high fidelity, the digitalization of high definition TV is one example of progress that has been made along the vertical axis. Signal processing technology is indispensable for the generation, processing, and recognition of visual, auditory and control signals. Furthermore, multimedia devices are preferable to work in real-time. These requirements can be fulfilled not only by high computing processor speeds and large memories, but also by fast algorithms. Therefore, the study of signal processing is fundamental in our laboratory.
Refereed Journal Papers
This new method employs a hand model and static images to analyze hand posture. Guided by internal constraints and external forces, the model is automatically fitted to the hand image.
The nearest neighbor multilayer perceptron (NN-MLP) is a single-hidden-layer network suitable for pattern recognition. To design an NN-MLP efficiently, this paper proposes a new evolutionary algorithm consisting of four basic operations: recognition, remembrance, reduction and review. Experimental results show that this algorithm can produce the smallest or nearly smallest networks from random initial ones.
The nearest neighbor based multilayer perceptron (NN-MLP) is a suitable model for self-organization, and has been studied by many authors in different forms. However, a large number of neurons are usually required in this kind of networks. To obtain smaller or the smallest NN-MLP, this paper introduces the concept of individual evolutionary algorithm (IEA), and proposes a new method for NN-MLP learning. There are four basic operations in the IEA: competition, gain, loss and retraining. The basic rule is: all individuals compete for surviving, winners gain more, losers lose more, and the individuals are retrained to function better than before. The learning algorithm based on the IEA is simple and suitable for parallel realization, and is able to produce the ``smallest-at-present'' networks from random ones in an evolutionary manner. Its efficiency is shown by experimental results.
This paper introduces a hairy brush modeling which enables users to draw ink-painting by controlling the 3D brush model on the paper. Elastic bristles are supposed to be fixed at the bottom of the holder to draw strokes. Basic strokes are generated and complex pictures are composed from them. Ink-painting by this model is proved to provide more naturalness, direct manipulation, and smaller complexity in drawing than conventional 2D outline-based methods.
A society S(I,T) is defined as a system consisting of two nonempty sets: an individual set I and a task set T. The problem is to find an efficient S such that all tasks in T can be fulfilled using the smallest I. If there is only one task, the best individual can be obtained by existing evolutionary algorithms. For multi-task problems, this paper proposes a new evolutionary algorithm which can produce the best individual set by successively performing four operations: competition, gain, loss and retraining. To make the study more concrete, the new algorithm is applied to learning of nearest neighbor based multilayer perceptrons, and experimental results show that the smallest or nearly smallest networks can be obtained.
A society S(I,T) is defined as a system consisting of an individual set I and a task set T. This paper studies the problem to find an efficient S such that all tasks in T can be fulfilled using the smallest I. The individual evolutionary algorithm (IEA) is proposed to solve this problem. By IEA, each individual finds and adapts itself to a class of tasks through evolution, and an efficient S can be obtained automatically. The IEA consists of four operations: competition, gain, loss and retraining. Competition tests the performance of the recent I and the fitness of each individual; gain increases the performance of I by adding new individuals; loss makes I more compact by removing individuals with very low fitness; and individuals are adjusted by retraining to make them better. An evolution cycle is: competition and (gain or loss) and retraining, and the evolution is performed cycle after cycle until some criterion is satisfied. The performance of IEA is verified by applying it to the learning of nearest neighbor based multilayer perceptrons.
The nearest neighbor based multilayer perceptron (NN-MLP) is a one-hidden-layer network suitable for pattern recognition. To get the smallest NN-MLP on-line, this paper introduces an evolutionary algorithm consisting of four operations: recognition, remembrance, reduction and review. The operation recognition tests the ability of the network and the fitness of each hidden neuron; the operation remembrance creates some new hidden neurons when the network functions too poor; the operation reduction removes some hidden neurons with very low fitness; and the network is readjusted by the operation review to achieve better performance. A learning cycle is defined as: recognition and (remembrance or reduction) and review, and the learning is performed cycle after cycle until some criterion is satisfied. Efficiency of the new algorithm is shown by experimental results.
The network considered in this paper is a three layered perceptron for realization of the fuzzified multi-template-matching. To find the smallest network with high generalization ability, this paper introduces the individual evolutionary algorithm (IEA), which was proposed by the author earlier for solving multi-task evolutionary problems. Experimental results show that, for off-line learning, IEA can produce the smallest or nearly smallest networks.