Momoyo Ito

Implementation of EOG mouse using Learning Vector Quantization and EOG-feature based methods

It is difficult for patients with severe physical disabilities to communicate with others, such as amyotrophic lateral sclerosis and serious paraplegia. Owing to the illness in which they lost their limb motor function and language function, they cannot move even their muscles except eye. In order to provide an efficient means of communication for those patients, in this paper we proposed a system that uses EOG-feature based methods and Learning Vector Quantization algorithm to recognize eye motions. According to the recognition results, we use API (application programming interface) to control cursor movements. The recognition part consists of four steps. First, we measure EOG signals by every 1.8 seconds. Next, we make a judge whether eye motion subsists in the 1.8 seconds EOG data, if any, we extract the data of each motion from the 1.8 seconds EOG data. After that we use Fast Fourier Transform to obtain the frequency features of the extracted motion. Finally we use Learning Vector Quantization network and characteristics of EOG features at each motion to recognize eye motions. The LVQ network is trained beforehand. In this paper we recognized motions of rolling eye upward, rolling downward, rolling left, rolling right, blink and diagonal eye motions which contain rolling up-left, rolling up-right, rolling down-left, rolling down-right (the angle of the diagonal motion is 45°) and blink string of three times motion. 8 directions motions correspond to 8 directions cursor movement in this system. We regard blink motion as invalid signal and define blink string motions as double click action. Using this system we have obtained a high recognition accuracy of eye motions (The average correct detection rate on each subject was 97.8%, 97.6% and 92.7%). This EOG Mouse interface would be used as a means of communication to help those patients as ALS.

Operation improvement of indoor robot by gesture recognition

Recently, the demand for the indoor robots has increased. Therefore, increased opportunities for many people to operate the robots have emerged. However, for many people, it is often difficult to operate a robot using the conventional methods like remote control. To solve this problem, we propose a robot operation system using the hand gesture recognition. Our method pays attention to the direction and movement of the hand. We were able to recognize several gestures in real-time.

Feature selection method for music mood score detection

In general, music retrieval and classification methods using music moods use a lot of acoustic features similar to music genre classification. These features are used as the spectral features, the rhythm features, the harmony features, and so on. However, all of these features may not be efficient for music retrieval and classification using music moods. Hence, in this paper, we propose a feature selection method for detecting music mood scores. In the proposed method, features which have strong correlation with mood scores are selected from a lot of features. Then, these are input into Multi-Layer Neural Networks (MLNNs) and mood scores are detected every mood labels. For evaluating the proposed method, we conducted the music mood score detection experiments. Experimental results show that the proposed method improves the detection performance compared to not use the feature selection.

Unsupervised segmentation for MR brain images

As described herein, we propose an unsupervised method for segmentation of magnetic resonance (MR) brain images by hybridizing the self-mapping characteristics of 1-D Self-Organizing Maps (SOMs) and using incremental learning functions of fuzzy Adaptive Resonance Theory (ART). As the proposed method requires the appropriate parameters to segment tissues (such as cerebrospinal fluid, gray matter and white matter) that are necessary for brain atrophy diagnosis, first we derive the optimal parameter set through the preliminary experiments. The main contribution of this work is to evaluate the effectiveness of the proposed method, considering the conventional methods that are highly accurate in terms of usefulness as classification techniques. We focus on Fuzzy C-means (FCM) and Expectation Maximization Gaussian Mixture (EM-GM) with previous setting of the number of clusters, and then Mean Shift (MS) without previous setting of the number of clusters. Through the comparative experiments on the two metrics, we confirmed that our method could achieve higher accuracy than these conventional methods. Additionally, we propose a Computer-Aided Diagnosis (CAD) system for use with brain dock examinations based on case analyses of diagnostic reading. We construct a prototype system for reducing loads on diagnosticians during quantitative analysis of the degree of brain atrophy. Field tests of 193 examples of brain dock medical examinees reveal that the system efficiently supports diagnostic work in the clinical field: the alteration of brain atrophy attributable to aging can be quantified easily, irrespective of the diagnostician.

Development of automatic filtering system for individually unpleasant data detected by pupil-size change

We proposed an automatic filtering system to classify individual unpleasant emotions represented by pupil-size change and to remove similar images from a multimedia database. The support vector machines classifier was applied to single-trial data of the pupil size and indicated the possibility of the accurate judgment of individually unpleasant states immediately after looking at emotional pictures. We then constructed the framework to automatically filter such unpleasant information from a picture database, using the bag of features scheme to search for similar images.

Novel supervised feature extraction algorithm based on iterative calculations

In pattern recognition, the principal component analysis (PCA) is one of the most famous feature extraction methods for dimensionality reduction of high-dimensional datasets. Furthermore, Simple-PCA (SPCA) which is a faster version of the PCA, has been carried out effectively by iterative operated learning. However, in SPCA, when input data are distributed in a complex way, SPCA might not be efficient because it is learned without class information of the dataset. Thus, SPCA cannot be said that it is optimal for classification. In this paper, we propose a new learning algorithm, which is learned with the class information of the dataset. Eigenvectors spanning eigenspace of the dataset are obtained by calculation of data variations belonging to each class. We will show the derivation of the proposed algorithm and demonstrate some experiments to compare the SPCA with the proposed algorithm by using UCI datasets.

Brain tissues segmentation for diagnosis of Alzheimer-type Dementia

We proposed a segmentation method of brain tissues on T2-weighted MR frontal image. In our previous work, we showed an image diagnosis support system for Alzheimer-type Dementia that extracts temporal lobe regions and an intracranial region as regions of interest (ROI) from a T2-weighted MR frontal image and uses the cerebral atrophy rates at the ROI. In this paper, we specifically discuss segmentation of brain tissues which are used for calculation of atrophy rate. We proposed a brain tissue segmentation method using two kinds of unsupervised neural networks: Self-Organizing Maps (SOMs) and Fuzzy Adaptive Resonance Theory (ART). The performance of proposed method was tested in two brain MR images used in daily diagnosis. Proposed method could segment CSF accurately with continuity of brain tissues.

Improvement in detection of abandoned object by pan-tilt camera

Recently, security cameras have been installed at a high rate in places where there are extensive grounds and many humans gather. The number of installed security cameras has been increasing year by year. The main reason is security enhancement including the prevention of incidences of terrorism. Therefore, we propose a method which detects abandoned objects on online by using pan-tilt camera. Above all, we improve problems of the previous method which is based on ST-Patch features and human detection. We make extended ST-Patch features for solving the problem of ST-Patch features. We improve human detection by using deep learning which is based on a convolutional neural network. We conducted preliminary experiments to verify a method of pooling, and then we decided to use Max pooling because its detection accuracy is better than that of Ave pooling. We conducted experiments in five situations to verify usefulness of the proposed method. If the proposed method finds an abandoned object, it saves the object image. We define the abandoned object as an object which human does not subsist near. We could detect the abandoned object in each situation. However, we conducted experiments of the proposed method only in a room. We need to conduct experiments in a wide area to find new problem.

Abandoned object detection by genetic algorithm with local search

In this study, we propose a method in which pictures of security cameras are administered automatically. The administered target is abandoned objects. In case of searching objects with security camera, there are infinitely various sizes and orientations of the object to be searched. Therefore, we propose an object search method which is adapted to transformation of the object. We use a template matching using Genetic Algorithm (GA) for detection of abandoned objects. Moreover, GA is suitable for global problems, but it is not necessarily suitable for local problems. Therefore the local search technique is included to improve GA property. Object search in our proposed method is divided into two parts: global search and local search. In the local search, we use a simple random search. According to experimental results, detection accuracy is relatively good in the global domain search, but the local domain search is no so effective in some images. In future work, we try to improve the local search.

An analytical method for face detection based on image patterns of EEG signals in the time-frequency domain

Although face-to-face communication includes the richest information, amyotrophic lateral sclerosis patients cannot smoothly communicate with others and express their emotions because of paralyzed muscles. Therefore, the N170 responses of EEG signals were analyzed to detect face stimuli in real time. We also proposed an analytical method for feature extraction of a support vector machine (SVM) classifier with the bag of features scheme to overcome the general difficulty in setting of kernel parameters of SVM. The proposed method resulted in a constantly high accuracy in the face classification; the SVM classifier based on image pattern recognition in the time frequency domain efficiently enables easier setting of the non-linear kernel parameter. Further studies will be required to apply the proposed method for feature extraction to practical devices.