This paper concerns the dynamical behavior, in probabilistic sense, of a feedforward neural network performing auto association for novelty. Networks of retinotopic topology having a one-to-one correspondence between and output units can be readily trained using back-propagation algorithm, to perform autoassociative mappings. A novelty filter is obtained by subtracting the network output from the input vector. Then the presentation of a "familiar" pattern tends to evoke a null response ; but any anomalous component is enhanced. Such a behavior exhibits a promising feature for enhancement of weak signals in additive noise. As an analysis of the novelty filtering, this paper shows that the probability density function of the weigh converges to Gaussian when the input time series is statistically characterized by nonsymmetrical probability density functions. After output units are locally linearized, the recursive relation for updating the weight of the neural network is converted into a first-order random differential equation. Based on this equation it is shown that the probability density function of the weight satisfies the Fokker-Planck equation. By solving the Fokker-Planck equation, it is found that the weight is Gaussian distributed with time dependent mean and variance.

For large vocabulary speech recognition using HMMs, context-dependent subword units have been often employed. However, when context-dependent phone models are used, they result in a system which has too may parameters to train. The problem of too many parameters and too little training data is absolutely crucial in the design of a statistical speech recognizer. Furthermore, when building large vocabulary speech recognition systems, unseen triphone problem is unavoidable. In this paper, we propose the modified phonetic decision tree algorithm for the automatic prediction of unseen triphones which has advantages solving these problems through following two experiments in Japanese contexts. The baseline experimental results show that the modified tree based clustering algorithm is effective for clustering and reducing the number of states without any degradation in performance. The task experimental results show that our proposed algorithm also has the advantage of providing a automatic prediction of unseen triphones.

This paper presents an implementation of a multi-channel speech surveillance system over telephone lines using TMS320C31 DSP chips. The incoming speech into each telephone line are first compressed simultaneously in real-time by the popular vector-sum excited linear predictive (VSELP) speech coding algorithm at the rate of 8 Kbps. The compressed steech bit streams are then multiplexed with those of other users. The multiplexed speech bit streams are transferred to the system storage equipments with some other required information so that a system operator can later monitor the stored speech data whenever it is necessary. The host program runs under Microsoft Windows95 for an efficient man-machine interface and a future upgrade-ability. We have confirmed that the overall 64-channel system operates satisfactorily in realtime. We also have checked approximately up to 2,880 total hours of recording capability of the system on a playback module and two removable backup drives.

A modified perturbation method is proposed to optimize the beam pattern of thinned microphone arrays. Both microphone spacing and array weight are iteratively adjusted via successive perturbation to achieve an optimum beam pattern in a Dolph Chebyshev sense. To improve the sidelobe performance, an alternative perturbation with respect to microphone spacing and array weight is implemented. Also, a linear space-tapering is employed in the perturbation process. It is demonstrated that the proposed approaches successfully yield sidelobe performances comparable to that of a normal array. Computer simulation results are presented.

In most switching system, the processing unit is designed to work efficiently even at relatively high loads, but when the offered traffic exceeds a particular level, the rate of completed calls can fall drastically. A single call handled by the switching system consists of a sequence of events or messages that has to be processed by the control unit. The control unit is not only incapable of handling all of the offered calls, but also its call handling capability can drop as the offered load increases. The real time available for call processing is a critical resource that requires careful management. Therefore, the overloading of this resource must be detected by a subscriber in the from of a dial tone delay or an uncompleted call which is either blocked or mishandled. The subscriber may respond by either dialing prematurely or by re-attempting a call. This action can further escalate the processors load, which is spent for uncompleted calls. Unless a proper control is used, the switching system can finally break down. In this paper, we paper, we propose a fuzzy overload detection and control method for switching systems, which can by generating fuzzy rules via fuzzy aggregation networks. Simulation results involving a switching system is given.

In this paper, an efficient quantization method of line spectrum pairs(LSP) with cascaded structure of vector quantizer and scalar quantizer is proposed. First, input LSP parameters is vector-quantized using a codebook a with a moderate number of entries. In the second stage of quantization, the components of residual vector are individually quantized by the scalar quantizer. The utilization of ordering property of LSP parameters and the inclusion of interframe prediction improve the quantizer performance and remove the stability check routine after quantization procedure. The new vector-scalar hybrid quantizer using 26 bits/frame shows a transparent quality of speech that an average spectral distortion is 1 dB and the frame proportion with above 2 dB spectral distortion is less than 2%. The performances of proposed quantization method is evaluated in the transmission errors.

Microphone is the first link in the speech recognition system. Depending on its type and mounting position, the microphone can significantly distort the spectrum and affect the performance of the speech recognition system. In this paper, characteristics of the speech signal for different microphones and microphone distances are investigated both in time and frequency domains. In the time domain analysis, the average signal-to-noise ration is measure ration is measured for the database we collected depending on the microphones and microphone distances. Mel-frequency spectral coefficients and mel-frequency cepstrum are computed to examine the spectral characteristics. Analysis results are discussed with our findings, and the result of recognition experiments is given.