A log-linearized Gaussian mixture network and its application toEEG pattern classification
IEEE Transactions on Systems, Man, and Cybernetics, Part C: Applications and Reviews
| Hi-index | 0.00 |
Many motion support robots of the elderly and disable were studies all over the world. We have developed the rollover support system, which is one of the ADL. Our ultimate goal is to develop an effective rollover support system for patients with cancer bone metastasis. The core of this system is a pneumatic rubber muscle that is operated by EMG signals from the trunk muscle. A Time Delay Neural Network (TDNN) is the traditional method for recognizing EMG signals. However, response delay and false recognition are the problem of the traditional neural network. We previously proposed a new neural network, called the Micro-Macro Neural Network (MMNN), to recognize the rollover movement earlier and with more accuracy than is possible with TDNN. MMNN is composed of a Micro Part, which detects rapid changes in the strength of the EMG signal, and a Macro Part, which detects the tendency of the EMG signal to continually increase or decrease. However, the methodology to determine the structure of the MMNN was not established. In this paper, the optimal structure of the MMNN is determined. A comparison of each of the 360 sets of test times of MMNN versus TDNN was done. These results showed that recognition using MMNN is 40 (msec) (S.D. 49) faster than recognition using TDNN. Additionally, the number of false recognitions using MMNN is one-third of that using TDNN. By comparing the output using only the Micro part and Macro part in MMNN, it was found that the combination of quick response of the Micro part and stable recognition of the Macro part are advantages of MMNN. In the future, we plan to test the effectiveness of the total system in clinical tests with cancer patients in terminal care.