Takatomi Kubo

Heart rate variability predicts the emotional state in dogs

Although it is known that heart rate variability (HRV) is a useful indicator of emotional states in animals, there are few reports of research in dogs. Thus, we investigated the relationship between HRV and emotional states in dogs. The electrocardiogram and behavior in two situations that elicited a positive and negative emotion, in addition to baseline (when dogs were not presented any social stimuli), were recorded in 33 healthy house dogs. After testing, we chose 15seconds from each situation and baseline and calculated three HRV parameters: standard deviation of normal-to-normal R-R intervals (SDNN), the root mean square of successive heartbeat interval differences (RMSSD), and mean R-R intervals (mean RRI). In comparing these parameters with baseline, only SDNN was lower in a positive situation. In contrast, only RMSSD was lower in a negative situation. A change in HRV occurred with a stimulus eliciting emotion, and was able to distinguish between positive and negative situations. Thus, HRV is useful for estimating the emotional state in dogs.

Art expertise reduces influence of visual salience on fixation in viewing abstract-paintings.

When viewing a painting, artists perceive more information from the painting on the basis of their experience and knowledge than art novices do. This difference can be reflected in eye scan paths during viewing of paintings. Distributions of scan paths of artists are different from those of novices even when the paintings contain no figurative object (i.e. abstract paintings). There are two possible explanations for this difference of scan paths. One is that artists have high sensitivity to high-level features such as textures and composition of colors and therefore their fixations are more driven by such features compared with novices. The other is that fixations of artists are more attracted by salient features than those of novices and the fixations are driven by low-level features. To test these, we measured eye fixations of artists and novices during the free viewing of various abstract paintings and compared the distribution of their fixations for each painting with a topological attentional map that quantifies the conspicuity of low-level features in the painting (i.e. saliency map). We found that the fixation distribution of artists was more distinguishable from the saliency map than that of novices. This difference indicates that fixations of artists are less driven by low-level features than those of novices. Our result suggests that artists may extract visual information from paintings based on high-level features. This ability of artists may be associated with artists’ deep aesthetic appreciation of paintings.

Modeling and Prediction of Driving Behaviors Using a Nonparametric Bayesian Method With AR Models

To develop a new generation advanced driver assistance system that avoids a dangerous condition in advance, we need to predict driving behaviors. Since a nonparametric Bayesian method with a two-level structure successfully predicted the symbolized behaviors only, we applied a nonparametric Bayesian method with linear dynamical systems to predicting the driving behavior. The method called the beta process autoregressive hidden Markov model (BP-AR-HMM) segments driving behaviors into states each of which corresponds to an AR model and it predicts future behaviors using the estimated future state sequence and the dynamical systems therein. Here, the segmentation as well as the parameters of the dynamical systems are determined using given training data in an unsupervised way. We carried out experiments with real driving data and found that the BP-AR-HMM predicted driving behaviors better than other methods.

Towards excluding redundancy in electrode grid for automatic speech recognition based on surface EMG

In our previous studies, an electrode grid was effective for Japanese vowel recognition from surface electromyography, and it was illustrated that a feature selection method compressing the features to one twentieth of the total features could be achieved without severe decline in recognition accuracies with one subject. In this study, we further verify the validity of the method with six subjects and explore more appropriate electrode locations without redundancy that can be generalizable over subjects. The results of this study indicate that feature selection can be realized while discriminative powers are kept to some extent for the all subjects. In addition, the channels in the central part of the electrode gird can be regarded as redundant with respect to some subjects. Thus, combining the dense measurements provided by the electrode grid and the feature selection method is an effective approach to explore appropriate measurement location without redundancy for the sEMG-ASR.

Towards prediction of driving behavior via basic pattern discovery with BP-AR-HMM

Prediction of driving behaviors is important problem in developing the next-generation driving support system. In order to take account of diverse driving situations, it is necessary to deal with multiple time series data considering commonalities and differences among them. In this paper we utilize the beta process autoregressive hidden Markov model (BP-AR-HMM) that can model multiple time series considering common and different features among them using the beta process as a prior distribution. We apply the BP-AR-HMM to actual driving behavior data to estimate VAR process parameters that represent the driving behaviors, and with the estimated parameters we predict the driving behaviors of unknown test data. The results suggest that it is possible to identify the dynamical behaviors of driving operations using BP-AR-HMM, and to predict driving behaviors in actual environment.

A comparative study of time series modeling for driving behavior towards prediction

Prediction of driving behaviors is an important problem in developing a next-generation driving support system. In order to take diverse driving situations into account, it is necessary to model multiple driving operation time series data. In this study we modeled multiple driving operation time series with four modeling methods including beta process autoregressive hidden Markov model (BP-AR-HMM), which we used in our previous study. We quantitatively compared the modeling methods with respect to prediction accuracies, and concluded that BP-AR-HMM excelled the other modeling methods in modeling multiple driving operation time series and predicting unknown driving operations. The result suggests that BP-AR-HMM estimated behaviors of a driver and transition probabilities between the behaviors more successfully than the other methods, because BP-AR-HMM can deal with commonalities and differences among multiple time series, but the others cannot. Therefore BP-AR-HMM may help us to predict driver behaviors in real environment and to develop the next-generation driving support system.

Variational Bayesian Inference Algorithms for Infinite Relational Model of Network Data

Network data show the relationship among one kind of objects, such as social networks and hyperlinks on the Web. Many statistical models have been proposed for analyzing these data. For modeling cluster structures of networks, the infinite relational model (IRM) was proposed as a Bayesian nonparametric extension of the stochastic block model. In this brief, we derive the inference algorithms for the IRM of network data based on the variational Bayesian (VB) inference methods. After showing the standard VB inference, we derive the collapsed VB (CVB) inference and its variant called the zeroth-order CVB inference. We compared the performances of the inference algorithms using six real network datasets. The CVB inference outperformed the VB inference in most of the datasets, and the differences were especially larger in dense networks.

Feature selection for vowel recognition based on surface electromyography derived with multichannel electrode grid

This paper investigates how feature selection influences the accuracy of vowel recognition based on surface electromyography (sEMG) derived with an electrode grid, which consists of densely-spaced multielectrodes. In previous studies on sEMG-based automatic speech recognition (sEMG-ASR), disc electrodes or parallel bar electrodes were used and located empirically. But, in this study, to avoid missing out information about speech, an electrode grid was used to measure sEMG from the submental region during the production of five Japanese vowels. For feature selection, we applied sparse discriminant analysis (SDA) to the obtained data which can include some redundant signals. It was illustrated that feature selection compressing to one tenth or one twentieth of the total features could be achieved without steep decline in recognition accuracies. Combination of dense measurement based on the electrode grid and feature selection based on SDA is an effective approach for researches on sEMG-ASR.

Effective napping support system by hypnagogic time estimation based on heart rate sensor

In daily life, lack of sleep is one of the main reasons for poor concentration. To support an effective napping, considered as one of good methods for recovering insufficient sleep and enhancing a users concentration, we propose a hypnagogic time estimation using a heart rate sensor. Because a heart rate sensor has already been common, our method can be used widely and easily in our daily life. Most of existing sleep support systems aim to provide a comfortable wake-up by observing the sleep stage. Unlike these methods, we aim to provide an appropriate sleep duration by estimating a hypnagogic timing. By using various heart rate sensors, existing sleep support systems and 64ch electroencephalography, we tried to find out the relationship between various vital signals and sleep stages during a napping. Finally, we build a hypnagogic time estimation model by using the machine learning technique.

Roles of pre-training in deep neural networks from information theoretical perspective

Although deep learning shows high performance in pattern recognition and machine learning, the reasons remain unclarified. To tackle this problem, we calculated the information theoretical variables of the representations in the hidden layers and analyzed their relationship to the performance. We found that entropy and mutual information, both of which decrease in a different way as the layer deepens, are related to the generalization errors after fine-tuning. This suggests that the information theoretical variables might be a criterion for determining the number of layers in deep learning without fine-tuning that requires high computational loads.