Cognitive Computation and Systems最新文献

Music can be regarded as an art of expressing inner feelings. However, most of the existing networks for music generation ignore the analysis of its emotional expression. In this paper, we propose to synthesise music according to the specified emotion, and also integrate the internal structural characteristics of music into the generation process. Specifically, we embed the emotional labels along with music structure features as the conditional input and then investigate the GRU network for generating emotional music. In addition to the generator, we also design a novel perceptually optimised emotion classification model which aims for promoting the generated music close to the emotion expression of real music. In order to validate the effectiveness of the proposed framework, both the subjective and objective experiments are conducted to verify that our method can produce emotional music correlated to the specified emotion and music structures.

Energy consumption in buildings is a major contributor to global warming and therefore has become a field of intensive research. This type of energy consumption can be described in two dimensions: an appliance-based dimension and a behaviour-based dimension. To address the behaviour-based dimension a recent study proposed a cognitive human-building interaction model that builds on the instance-based learning paradigm. However, since the values of the standard cognitive parameters commonly used for modelling lab-based behaviours are not suitable for the ‘real-world’ domain of human-building interaction, this paper aims to identify cognitive parameter values adapted to and suitable for the specific character of this application domain. To achieve this goal, a virtual test environment—consisting of an occupied room and a corresponding model task—was designed to test the performance of the model and its dependence on a set of fundamental cognitive parameters. A test criterion was developed that did not depend on empirical data but used the predictive consistency of the model as reference. A range of values was pre-selected for each parameter based on theoretical and empirical considerations, which was then tested against the evaluation criterion. The performance of the model was improved significantly throughout the parametrisation process and yielded plausible results.

The paper presents a novel algorithm to classify children's epileptic syndromes based on the fused features of electroencephalogram (EEG) and electrocardiogram (ECG). The purpose is to assess whether multimodal physiological signals could improve the classification performance of epileptic syndromes over a single physiological signal. The study is carried out on the epileptic syndromes database recorded by the Children's Hospital, Zhejiang University School of Medicine (CHZU), that includes the synchronised EEGs and ECGs of 16 children suffered from the infantile spasms (known as the WEST syndrome, named) and the childhood absence epilepsy (CAE), respectively. Experiments are conducted and compared using the EEGs and ECGs in the ictal and interictal periods. The data imbalanced issue between the ictal and interictal periods is also considered by applying a synthetic minority sample generating approach. The experimental results show that using the fused feature of EEG + ECG can achieve an average of 98.15% overall classification accuracy, which is better than using the single physiological signal.

The snubber circuit plays an important role in motor drives. This paper deals with the detection of the inverter switch snubber circuit resistance fault (ISSCRF) in brushless direct current (BLDC) motors used for robotic applications. This has been carried out in two parts: Fast-Fourier-Transform-based analysis and wavelet-decomposition-based analysis on the stator current of the BLDC motor. The first analysis investigates the effects of different percentages of ISSCRF on direct current (DC) component, fundamental frequency component and total harmonic distortion percentage. Next analyses consider all of kurtosis, skewness and root-mean-square values of wavelet coefficients of stator current harmonic spectra. Comparative learning is made to obtain a few selective parameters best fit for the detection of ISSCRF. A fault detection algorithm to detect ISSCRF has been proposed and validated by three case studies. The algorithm is again modified with best-fit parameters. Comparative discussion and novel contributions of the work have also been presented.

This paper details a solution of fusing combination features, Iterative Closest Point (ICP) and Monte Carlo algorithm, in order to solve the problem that mobile robot positioning is easy to fail in a dynamic environment. Firstly, an ICP algorithm based on the maximum common combination feature is proposed to provide a more stable observation point information and therefore avoids the problem of local extremes and obtains more accurate matching results. A novel proposal distribution is then designed and auxiliary particles are used, so that the particle sets are distributed in high-observational areas closer to the true posterior probability of the state. Finally, the experimental results on the public datasets show that the proposed algorithm is more accurate in these environments.

Flexible tactile sensing based on capacitive sensing has become a research hotspot in recent years because of its low energy consumption, high performance and wide application prospects. However, the axis error caused by the coupling deformation of the dielectric will seriously affect the accuracy of the sensor. In this paper, a capacitive flexible three-axis tactile sensor array is modelled and simulated, and a neural network-based calibrator for the three-axis sensor array is proposed, which can be used to calibrate the simulated measurement data. The simulation results show that even though the correlation coefficient of linear regression for each axis is very close to 1, the effect of dielectric nonlinear coupling distortion cannot be eliminated. The calibration method based on the neural network can effectively suppress the nonlinear coupling distortion of the dielectric, and reduce the measurement coupling rate of the sensor model from 26% to 1%. At the same time, in order to ensure the measurement accuracy and robustness of different units in the sensor array, the input layer of the calibrator is expanded, and the data set containing capacitance information and two-dimensional location information is used for training. The experimental results show that the proposed calibration method combining two-dimensional position information training accurately calibrates the capacitive flexible three-dimensional tactile sensor array.

In a cross-cultural context, exploring musical elements' cultural specificity and universality that affect various types of music is conducive to personalised emotion recognition. In this study, high-level musical elements are introduced to explore their influence on emotional perception. By comparing music emotion recognition (MER) models of varied cultural music, musical elements with cultural universality and cultural specificity are further determined. Participants rated valence, tension arousal, and energy arousal on labelled nine-point analogical–categorical scales for four types of classical music: Chinese ensemble, Chinese solo, Western ensemble, and Western solo. Fifteen musical elements in five categories—timbre, rhythm, articulation, dynamics, and register were annotated through manual evaluation or the automatic algorithm. The relationship between music emotion and musical elements was analysed through partial least squares regression. Results showed that tempo, rhythm complexity, and articulation are culturally universal; musical elements related to timbre, register, and dynamics features are culturally specific. By increasing tempo, rhythm complexity, staccato, perception of valence, tension arousal, and energy arousal can be effectively improved. Based on the Partial least squares regression (PLSR) model's results for the datasets, the combination of manual and automatic annotation for musical elements can improve the MER system's performance.

Gas-insulated switchgear (GIS) is an important power equipment. The implementation of health monitoring is limited by the number of sensors, and the global detection results of the system should be highly credible to ensure the reliability of the power supply system. To solve this problem, this study proposes a sensor layout optimization method based on global detection probability performance evaluation. Starting from the cost function, the GIS discharge detection problem is transformed into a Bayesian risk decision problem, the binary state of ‘with discharge’ and ‘without discharge’ is adopted to simplify the cost function and reduce the computing workload, and the objective function representing the global detection performance of the system is obtained. The solution of layout optimization is realized by the improved genetic algorithm. 3-sensor, 4-sensor and 6-sensor layouts, which are digitally simulated at different detection rates, and then the distribution diagram of the global detection rate is obtained. On this basis, the feasibility and effectiveness of the optimization method are verified through an experiment. The results show that, compared with other sensor layout optimization methods, this optimization method can obtain the correct probability distribution of the detection rate globally and realize the graphical quantization of the detection performance distribution of the system so as to ensure the system performance.

In order to solve the problem of low accuracy and efficiency of soil moisture content prediction in tea plantations and improve the level of soil water content prediction, a soil moisture content prediction model for tea plantations based on the support vector machine (SVM)-optimised bald eagle search (BES) algorithm (BES-SVM) is proposed. Soil data and environmental data of tea plantations were transmitted to the server using sensor nodes and weather station nodes. The prediction models of soil moisture content and natural environmental parameters such as soil electrical conductivity, soil temperature, air temperature, air humidity, light intensity, and rainfall were developed using the SVM model optimised by the bald eagle search algorithm, and the mean square error (MSE) and coefficient of determination ($��R^2�$) were calculated to evaluate the model performance. Meanwhile, the performance of the BES-SVM model is compared with the particle swarm algorithm optimisation SVM (PSO-SVM) and genetic algorithm optimised SVM (GA-SVM) models. Results show that the proposed model has a mean coefficient of determination of 95.65%, and the prediction performance is better than the PSO-SVM and GA-SVM model, indicating that the BES-SVM model has good performance and is a feasible prediction method for soil water content prediction and guiding irrigation and fertilisation management in tea plantations.

This is the Special Issue ‘Integrating Sensor Fusion and Perception for Human–Robot Interaction’ of IET Cognitive Computation and System that introduces the latest advances in sensor fusion and perception in the human–robot interaction (HRI) field.

In recent years, as intelligent systems have developed, HRI has attracted increasing research interest. In many areas, including factories, rehabilitation robots and operating rooms, HRI technology can be exploited to enhance safety by using intelligence for human operations. However, both available practical robotic systems and some ongoing investigations lack intelligence due to their limited capabilities in perceiving their environment. Nowadays, the HRI method usually focusses on a single sensing system without integrating algorithms and hardware, such as tactile perception and computer vision. Sensor fusion and perception with artificial intelligence (AI) techniques have been successful in environment perception and activity recognition by fusing information from a multi-modal sensing system and selecting the most appropriate information to perceive the activity or environment. Consequently, combining the technique of multi-sensor fusion and perception for HRI is an exciting and promising topic.

This Special Issue aims to track the latest advances and newly appeared technology in the integrated sensor fusion and perception for HRI. After careful peer reviews and revision, four representative papers were accepted for publication in this Special Issue. These papers represent four important application areas of multi-sensor fusion and perception technology and can be assigned into four topics. The related summary of every topic is given below. We strongly recommend reading the entire paper if interested. They will bring some new ideas and inspire the mind.

In the paper ‘Deep learning techniques-based perfection of multi-sensor fusion oriented human-robot interaction system for identification of dense organisms’, Li et al. present an HRI system based on deep learning and sensors' fusion to study the species and density of dense organisms in the deep-sea hydrothermal vent. In this paper, several deep learning models based on convolutional neural network (CNN) are improved and compared to study the species and density of dense organisms in deep-sea hydrothermal vent, which are fused with related environmental information provided by position sensors and conductivity–temperature–depth (CTD) sensors, so as to perfect the multi-sensor fusion-oriented HRI system. First, the authors combined different meta-architectures and different feature extractors and obtained five object identification algorithms based on CNN. Then, they compared the computational cost of feature extractors and weighed the pros and cons of each algorithm from mean detection speed, correlation coefficient and mean class-specific confidence score to confirm that Faster Region-based CNN (R-CNN)_InceptionNet is