This paper proposes a method to estimate the waveform of a specified sound source in a noisy and reverberant environment using a sensor array. Previously, we proposed an iterative method to estimate the waveform. However, in this method the effect of reflection sound reduces to 1/M, where M is the number of microphones. Therefore, to solve the reverberation problem, we propose a new method using inverse filters of the transfer functions from the sound sources to each microphone. First, the transfer function from each sound source to each microphone is measured by the cross-spectrum technique and each inverse filter is calculated by the QR method. Then the initially estimated waveform of a sound source is the averaged signal of the inverse filter outputs. Since this waveform still contains the effects of the other sound sources, the iterative technique is adopted to estimate the waveform more precisely, reducing the effects of the other sound and the reflection sound. Some computer simulations and experiments were carried out. The results show the effectiveness of our method.

A non-invasive method is proposed to estimate the location of intracranial vascular disease using several sensors placed on the forehead. The advantage of this method over earlier measurements with a single ocular sensor is the abilty to localize the region of abnormal vascular tissue. A weighted least mean square procedure is applied to estimating the time difference between the sensor outputs using the phase distribution in the cross-spectrum. It is possible to estimate time differences shorter than sampling period. Computer simulation and clinical experiments demonstrate that a distance difference of around 20 times shorter than the wavelength can be obtained.

In earlier works we proposed the Exclusive Learning neural NET work (ELNET), which can be utilized to construct large scale recognition system for Chinese characters. However, this did not resolve the problem of how to use training samples effectively to generate more suitable recognition boundaries. In this paper, we propose ELNET- wherein an attempt has been made to deal with this problem. In comparison with ELNET, selection method of training samples is improved. And the number of module size are variable according to the number of training samples for each module. In recognition experiment for ETL9B (3036 categories) using ELNET-, we obtained a recognition rate of 95.84% as maximum recognition rate. This is the first time that such a high recognition rate has been obtained by neural networks.