Search

This paper presents a new algorithm for incremental learning, which is named Incremental Simple-PCA. This algorithm adds an incremental learning function to the Simple-PCA that is an approximation algorithm of the principal component analysis where an eigenvector can be calculated by a simple repeated calculation. Using the proposed algorithm, it is possible to update the eigenvector faster by using incremental data. We carry out computer simulations on personal authentication that uses face images and wrist motion recognition that uses wrist EMG by incremental learning to verify the effectiveness of this algorithm. These results were compared with the results of Incremental PCA that introduced incremental learning function to the conventional PCA.

This paper presents a new algorithm for feature generation, which is derived based on geometrical interpretation of the fisher linear discriminant analysis (FLDA). This algorithm (Simple-FLDA) is an approximation algorithm that calculates eigenvectors sequentially by an easy iterative calculation by expressing the maximization of variance between classes and minimization of variance in each class without the use of matrix calculation. We carry out computer simulations about recognition of wrist motion patterns by EMG measured from wrist and personal authentications that use face images to verify the effectiveness of this technique. The result was compared with the result of principal component analysis (Simple-PCA).

In recent years, ICT devices such as personal computer and video camera, tablet devices have been introduced actively in the field of ball sports. As one example, before the game is also carried out analysis of the opposing team using recorded video. In this paper, we consider the method to extract the desired scenes from the game video automatically for coaching support in the rugby sports. HOG and SVM were combined and verified whether human detection and scene extraction were possible from the actual game video. The experimental results, extraction of persons could be performed. However, it has been understood that the further verification is necessary for the scene extraction.

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.

There is electromyogram(EMG) as one of biological signals generated along with motions of the human body. This EMG has information corresponding to condition of power, tenderness of motions and motion level. Therefore, it is thought that it is useful biological information for analyzing person’s motions. Recently, researches on this EMG have been actively done. For instance, it is used as a control signal of electrical arms, because EMG can be gathered from the remaining muscle about upper-extremity amputee. In addition, the pointing device that uses EMG has been developed. In general, EMG is measured from a part with comparatively big muscular fibers such as arms and shoulders. There is a problem that placing and removing the electrode is inconvenient when EMG is measured by placing the electrode in the arm and the shoulder. Therefore, if we can recognize wrist motions using EMG which is measured from the wrist, the range of application will extend furthermore. Currently, we have constructed a wrist motion recognition system which recognizes the wrist motion of 7 types as an object by using the technique named Simple-FLDA as a feature extraction technique. However, motions with low recognition accuracy were observed, and it was found that the difference of the recognition accuracy is significant at each motion and subject. Because EMG is highly individual and its repeatability is low. Therefore, it is necessary to deal with these problems. In this paper, we try the construction of a system that can learn by giving incremental data to achieve an online tuning. The improvement of the algorithm of Simple-FLDA that incremental learning becomes possible was tried as a technique for the construction of online tuning system. As a recognition experimental result, we can confirm the rising of the recognition accuracy by incremental learning.

Recently, studies of artificial arms and pointing devices using ElectroMyoGram(EMG) have been actively done. However, the individual variation of EMG is large, and its repeatability is low. Furthermore, EMG is usually measured from a part with comparatively big muscular fibers such as arms and shoulders. Therefore, if we can recognize wrist operations using EMG which was measured from the wrist, the range of application will extend furthermore. In this study, we aim toward the development of a device of wristwatch type that consolidates operational interface of various equipments. In particular, as an early stage, we propose a wrist motion recognition system. First, we execute the Fourier transform to the signal for feature extraction. Next, we experiment it by using neural networks after the dimensional reduction by using Simple-PCA and Simple-FLDA to reduce the number of inputs to NN. It was confirmed that the present approach was one of the techniques which were effective in the wrist recognition experiment.