Search

The ageing of the population drives the rapid increase of cognitive disorders, which cause a heavy burden for families and nations. It is important to screen and interfere with cognitive disorders at the earlier stage, but we are short of affordable and effective approaches. In this talk, we will discuss our attempt in the design of the multi-mode detection and rehabilitation methods for cognitive disorders using computational intelligence algorithms, wearable devices and rehabilitation robots.

This paper presents a pattern recognition model using multiple features and the kernel extreme learning machine (ELM), improving the accuracy of automatic epilepsy diagnosis. After simple preprocessing, temporal- and wavelet-based features are extracted from epileptic EEG signals. A combined kernel-function-based ELM approach is then proposed for feature classification. To further reduce the computation, Cholesky decomposition is introduced during the process of calculating the output weights. The experimental results show that the proposed method can achieve satisfactory accuracy with less computation time.

Bonnet polishing has been successfully used for ultra-precision machining of the large-aperture aspheric optics. To accomplish online monitoring of executive condition of the removing function and state of the polishing machine, key technological parameters and processing condition factors need to be acquired in real time and coupled back. A novel data fusion method which combines hardware synchronous signal and correcting algorithm based on the principle of linear extrapolation is proposed in allusion to data acquisition distortion caused by asynchronous sampling frequency and communication delay between different systems. The estimate values are obtained by dynamically correcting original acquired data, and the upper bound of the error is deduced with the speed and the acceleration of the signal. The simulation results showed that the data fusion method can significantly reduce the error of data acquisition. The experimental results confirmed that the data fusion method can be effectively applied to monitor the optics bonnet polishing process online.

In neuro-rehabilitation after stroke, the conventional constrained induced movement therapy (CIMT) has been well-accepted. Existing bilateral trainings are mostly on mirrored symmetrical motion. However, complementary bilateral movements are dominantly involved in activities of daily living (ADLs), and functional bilateral therapies may bring better skill transfer from trainings to daily life. Neurophysiological evidence is also growing. In this work, we firstly introduce our bilateral arm training system realized with a haptic interface and a motion sensor, as well as the tasks that have been designed to train both the manipulation function of the paretic arm and coordination of bilateral upper limbs. Then, we propose quantitative measures for functional assessment of complementary bilateral training performance, including kinematic behavior indices, smoothness, submovement and bimanual coordination. After that, we describe the experiments with healthy subjects and the results with respect to these quantitative measures. Feasibility and sensitivity of the proposed indices were evaluated through comparison of unilateral and bilateral training outcomes. The proposed bilateral training system and tasks, as well as the quantitative measures, have been demonstrated effective for training and assessment of unilateral and bilateral arm functions.

Visual tracking is a popular and challenging topic in computer vision and robotics. Owing to changes in the appearance of the target and complicated variations that may occur in various scenes, online learning scheme is necessary for advanced visual tracking framework to adopt. This paper briefly introduces the challenges and applications of visual tracking and focuses on discussing the state-of-the-art online-learning-based tracking methods by category. We provide detail descriptions of representative methods in each category, and examine their pros and cons. Moreover, several most representative algorithms are implemented to provide quantitative reference. At last, we outline several trends for future visual tracking research.

It is widely believed that human brain is a complicated network and many neurological disorders such as Alzheimer's disease (AD) are related to abnormal changes of the brain network architecture. In this work, we present a kernel-based method to establish a network for each subject using mean cortical thickness, which we refer to hereafter as the individual's network. We construct individual networks for 83 subjects, including AD patients and normal controls (NC), which are taken from the Open Access Series of Imaging Studies database. The network edge features are used to make prediction of AD/NC through the sophisticated machine learning technology. As the number of edge features is much more than that of samples, feature selection is applied to avoid the adverse impact of high-dimensional data on the performance of classifier. We use a hybrid feature selection that combines filter and wrapper methods, and compare the performance of six different combinations of them. Finally, support vector machines are trained using the selected features. To obtain an unbiased evaluation of our method, we use a nested cross validation framework to choose the optimal hyper-parameters of classifier and evaluate the generalization of the method. We report the best accuracy of 90.4 % using the proposed method in the leave-one-out analysis, outperforming that using the raw cortical thickness data by more than 10 %.

Agent and Ontology are distinct technologies that arose independent of each other, having their own standards and specifications. The semantics web is one of the popular research areas these days, and is based on the current Web, which adds more semantics to it for the purpose of building the Ontology of Web content. In this regard, application program on Web can make the purpose of cross-platform calculation come true by taking advantage of Ontology. However, agent is a theory able to enhance abstraction of software itself, and as it is know, negotiation protocol is the basic principle in the electronic commerce which has a direct impact on the efficiency of the negotiation. This study examines the communication architecture with negotiation protocol on the Semantic Web. Precisely speaking, agents make computing with Ontology, and the authors define an agent’s communication ontology for this communication framework and semantic web use Ontology to describe the negotiation protocol. In this context, the buyer or seller will be able to improve semantic cognitive in process of negotiation. Also, it can provide an intelligent platform for the information exchange on the same understanding about the content of communication in the electronic negotiation service.

Multi-robot systems have been extensively studied and the coordination among the robots becomes a hotspot. Among the typical research tasks of multi-robot system, coordinated hunting with unknown irregular motion of the evader or target has attracted more and more attentions due to its potential application to military, safe guard etc. The spiking neural network (SNN), considered as the third generation of neural network (Maass & Bishop, 1999), has attracted many attentions. Spikes (pulses) are used to deliver the information between neurons, i.e. SNN processes the information in the form of spikes, which brings temporal structure and extends the functionality of SNN (Kasabov, 2010). Besides the rate coding, inspired by the results of biological experiments, some coding strategies based on spike timing have been proposed, such as time-to-first-spike coding, phase coding, correlations/synchrony coding et al. (Maass & Bishop, 1999). Also, Maass et al. present some useful spiking neuron models, such as spike response model (SRM), integrate-and-fire model (IF), Hodgkin-Huxley Model and so on (Maass & Bishop, 1999). For these coding methods and neuron models, the Hebb learning is proposed to adapt the weights between neurons based on the temporal difference between input and output spikes (Kempter et al., 1999). Nowadays, the controllers based on SNN have been introduced to many applications, such as phase/frequency correlations recognizing (Kiselev, 2009), movement prediction from real-world images (Burgsteiner et al., 2005), movement generation of the robot arm (Joshi & Maass, 2005), etc. Especially, SNN has been applied to the control of the mobile robot (Floreano & Mattiussi, 2001; Roggen et al., 2003; Hagras, 2004; Qu et al., 2009). In this chapter, a robot controller based on spiking neural network (SNN) is proposed for the coordinated hunting of multi-robot system. The controller utilizes 12 direction-sensitive modules to encode and process the inputs including the environment, target and coordination information by the time-to-first-spike coding and then, the motor neurons generate the control signals for the motors according to the winner-take-all strategy. Also, the Hebbian learning with a stochastic form is applied to adjust the connection weights. The rest of the chapter is organized as follows. In Section 2, the structure of the controller is given and the inputs encoding, coordination between robots and motor neurons are

In this article we propose a framework for establishing individual structural networks. An individual network is established for each subject using the mean cortical thickness of cortical regions as defined by the AAL atlas. Specifically, for each subject, we compute a similarity matrix of mean cortical thickness between pairs of cortical regions, which we refer to hereafter as the individual's network. Such individual networks can be used for classification. We use a combination of two types of feature selection approaches to search for the most discriminative edges. These edges serve as the input to a support vector machine (SVM) for classification. We demonstrate the utility of the proposed method by a comparison with classifying the raw cortical thickness data, and individual networks, using a publically available dataset. In particular, 83 subjects from the OASIS database were chosen to validate this approach, 39 of which were diagnosed with either mild cognitive impairment (MCI) or moderate Alzheimer's disease (AD) and the remaining were age-matched controls. While using an SVM on the raw cortical thickness data or individual networks without hybrid feature selection resulted in less than or nearly 80% classification accuracy, our approach yielded 90.4% classification accuracy in leave-one-out analysis.