In this paper, we propose the first study on the synthesis of the filter geometry directly from the given frequency response using a convolutional neural network (CNN). By assuming the planar filter geometry as an image, the CNN can learn the relationship between the geometry and frequency response because the CNN is good at dealing with image matter. We also explain a way to generate an accurate and massive dataset. The massiveness is achieved by high-speed lookup-table-based cascaded ABCD matrix calculation. The accuracy is guaranteed by the consideration of junctions where the line widths are different. As a result, the pair of a randomly generated filter geometry and its filter response is calculated in 2.0 ms. After the CNN is trained with the dataset containing 100k pairs, the CNN can synthesize a filter in 2.9 ms on an nVidia RTX 2080 graphics card. We also introduce a CNN that can estimate the frequency response directly from the filter geometry in 1.9 ms.

In this paper, we discuss distributed blind equalization based on a single-input multipleoutput (SIMO) channel model. To deal with a realistic situation, it is assumed that different channels make distributed blind equalization more difficult in a wireless sensor network (WSN). The performance is affected by the degree of severity of the noisy channel output. The eigenvalue spread of the input correlation matrix is utilized to propose a new combination rule whose coefficients are estimated from the neighboring channel outputs instead of only the degrees of the nodes. The eigenvalue-spread-based combination rule is implemented by two approaches, and compared with conventional combination rules by utilizing one of the distributed blind equalization algorithms. Mean square error (MSE) and symbol error rate (SER) are investigated for several communication channels. Computer simulations validate the superiority of the proposed combination rule to the conventional rules.

Spectrum sensing is an important issue in cognitive radio (CR) systems for solving spectrum scarcity problems in wireless communication systems. This paper presents a semiblind spectrum-sensing method utilizing higher order statistics including the skewness and kurtosis functions. The proposed method improves the detection performance while increasing its computational complexity, where an orthogonal frequency division multiplexing (OFDM)-transmitted signal over multipath fading channels is considered. Through comprehensive evaluation by simulations, it is shown that the proposed spectrum-sensing method significantly outperforms the conventional schemes for low signal-to-noise ratio (SNR) cases.

Blind image quality assessment (BIQA) methods can measure the quality of distorted images even without referencing the original images. This property is indispensable in the image processing field because reference images are normally not available in practice. Unlike the existing trained models, in our work, the training process is constructed as an end-to-end learning mechanism that minimizes the loss between the predicted score and the ground-truth score of the human vision system (HVS). Moreover, a convolutional neural network (CNN) takes distorted images as input and outputs the related score for each image. In this paper, we evaluate the proposed method on six publicly available benchmarks and the cross-database validation performance on the LIVE, CSIQ and TID2013 databases. The experimental results show that our proposed method outperforms other state-of-the-art methods.

Currently, many elderly people in Japan suffer from dysphagia (difficulty related to swallowing). The likelihood of aspiration of food owing to the state of the swallowing function has not yet been estimated. The change in the frequency domain for the classification of swallowing sounds ( I, II, and III sound) is confirmed in this study. The optimum number of microphones and their placement are examined by performing principal component analysis on the data gathered from a multi-channel microphone. Differences in frequency are found to confirm the classification of swallowing sounds. A higher frequency is generated in the latter part of III sound. Analysis in the frequency domain proved to be able to classify 80% of the signals. In addition, multi-channel measurement and analysis using principal component analysis is shown to be effective for feature extraction of swallowing sounds, which suggests that classification accuracy would be further improved by combining it with frequency analysis. In addition, it is suggested that by releasing the measurement position, it is possible to extract signals with different characteristics from each microphone.

Pneumonia is the main cause of death in Japan, with many deaths relate to aspiration pneumonia. However, there are no quantitative indicators for a noninvasive swallowing assessment. Typically, pharyngeal ultrasoundvideos (B-mode) are used as a noninvasive examination method. However, because determining an appropriate probe angle is difficult, the probe angle has typically been evaluated manually and empirically. In this study, a novel moving-image-processing method is evaluated for use in determining the probe angle. In each frame of both the moving and reference images, the similarity in discrete wavelet transform are calculated and discriminated. Results show that because the method analyzes similarity between the reference and original moving image, the Level 2 or 3 y-direction wavelet spaces are superior. As such, it is possible to determine an appropriate probe angle. Finally, to deal with individual differences, a matrix is created to determine the probe angle from the reference and moving images created from multiple subjects, and the reference images are selected.

This paper presents a method of pseudo-formal linearization based on the orthogonal polynomial approximation for nonlinear systems. The given nonlinear system is piecewise linearized by the formal linearization approach using the Legendre, Chebyshev, Hermite, or Laguerre polynomial functions. The resulting formal linear systems are smoothly united into a single linear one by an automatic choosing function. A nonlinear observer is synthesized as an application of this method. Numerical examples show that the accuracy of approximation by this method is greatly improved when appropriate orthogonal polynomial functions are chosen.