The paper introduces an electroencephalography U+0028 EEG U+0029 driven online position control scheme for a robot arm by utilizing motor imagery to activate and error related potential U+0028 ErrP U+0029 to stop the movement of the individual links, following a fixed U+0028 pre-defined U+0029 order of link selection. The right U+0028 left U+0029 hand motor imagery is used to turn a link clockwise U+0028 counterclockwise U+0029 and foot imagery is used to move a link forward. The occurrence of ErrP here indicates that the link under motion crosses the visually fixed target position, which usually is a plane U+002F line U+002F point depending on the desired transition of the link across 3D planes U+002F around 2D lines U+002F along 2D lines respectively. The imagined task about individual link U+00BC s movement is decoded by a classifier into three possible class labels: clockwise, counterclockwise and no movement in case of rotational movements and forward, backward and no movement in case of translational movements. One additional classifier is required to detect the occurrence of the ErrP signal, elicited due to visually inspired positional link error with reference to a geometrically selected target position. Wavelet coefficients and adaptive autoregressive parameters are extracted as features for motor imagery and ErrP signals respectively. Support vector machine classifiers are used to decode motor imagination and ErrP with high classification accuracy above 80 U+0025. The average time taken by the proposed scheme to decode and execute control intentions for the complete movement of three links of a robot is approximately 33 seconds. The steady-state error and peak overshoot of the proposed controller are experimentally obtained as 1.1 U+0025 and 4.6 U+0025 respectively.

Motor imagery and error related potential induced position control of a robotic arm

Discriminating metal parts of buried hazardous targets from ordinary metallic clutter is a very difficult and time-consuming task. For that purpose, electromagnetic induction (EMI) sensors composed of multiple transmitter/receiver coils in multiaxis arrangements are commonly used. In this way, data of high fidelity and spatial diversity are obtained so that the target can be characterized in terms of its geometric and electromagnetic properties. However, this often increases sensor size and reduces its portability, which is a problem for humanitarian demining applications, where compact, robust, and lightweight sensors are needed. If such sensors are to be used for target characterization, robust estimation algorithms are required, capable of coping with limited spatial information content and uncertainties related to sensor positioning and its coil geometry model. In this paper, we present a robust concept for estimating the general shape of magnetic metal targets using time-domain EMI sensors with single-axis coil geometries. We introduce the signature matrix, a parameter derived from time-dependent eigenvalues of the target’s magnetic polarizability tensor, and use it for shape estimation. The proposed method was evaluated both through simulations and experiments, using two different sensor platforms (laboratory-based experimental sensor platform and a commercial metal detector mounted on a mobile robot). The obtained results clearly indicate that the target shape can be estimated from sensor data of limited spatial diversity and under the uncertainties of sensor positioning and coil geometry.

Robust Estimation of Metal Target Shape Using Time-Domain Electromagnetic Induction Data

Frequency-domain electromagnetic induction (EMI) sensors can measure object-specific signatures that can be used to discriminate landmines from harmless clutter. In a model-based signal processing paradigm, the object signatures can often be decomposed into a weighted sum of parameterized basis functions, such as the discrete spectrum of relaxation frequencies (DSRF), where the basis functions are intrinsic to the object under consideration and the associated weights are a function of the target-sensor orientation. The basis function parameters can then be used as features for classifying the target. One of the challenges associated with effectively utilizing a model-based signal processing paradigm such as this is determining the correct model order for the measured data, as the number of basis functions containing fundamental information regarding the target under consideration is not known a priori. In this paper, sparse Bayesian relevance vector machine (RVM) regression is applied to simultaneously determine both the number of parameterized basis functions and their relative contributions to the measured signal assuming a DSRF signal model. The target is then classified utilizing the basis function parameters as features within a statistical classifier. Results for data measured with a prototype frequency-domain EMI sensor at a standardized test site are presented, and indicate that RVM regression followed by distance-based statistical classifiers utilizing the resulting model-based features provides an effective approach for classifying and identifying landmine targets.

Target Classification and Identification Using Sparse Model Representations of Frequency-Domain Electromagnetic Induction Sensor Data

Transient stability assessment (TSA) based on machine learning techniques is a classical problem of classification. In this paper, a new transient stability assessment model using the weighted random forest (WRF) is proposed. Random forest, which runs fast and can scale to very high dimensional datasets, is one of the most popular methods applied to TSA after faults occurred. However, in the actual operation of modern power system, unstable samples constitute only a very small minority of the cases. Besides, the traditional random forest algorithm does not work very well on imbalanced data. To tackle the problem, weighted random forest is adopted to lean the interest towards to the correct classification of rare class, and improve the prediction performance. Case study on the New England 39-bus test power system exhibits that WRF has better performance than the original random forest.

A weighted random forest approach to improve predictive performance for power system transient stability assessment

This paper presents an inversion procedure to estimate the location and magnetic polarizability tensor of metal targets from broadband electromagnetic induction (EMI) data. The solution of this inversion produces a spectral target signature, which may be used in identifying metal targets in landmines from harmless clutter. In this process, the response of the metal target is modelled with dipole moment and fitted to planar EMI data by solving a minimization least squares problem. A computer simulation platform has been developed using a modelled EMI sensor to produce synthetic data for inversion. The reconstructed tensor is compared with an assumed true solution estimated using a modelled tri-axial Helmholtz coil array. Using some test examples including a sphere which has a known analytical solution, results show the inversion routine produces accurate tensors to within 12% error of the true tensor. A good convergence rate is also demonstrated even when the target location is mis-estimated by a few centimeters. Having verified the inversion routine using finite element modelling, a swept frequency EMI experimental setup is used to compute tensors for a set of test samples representing examples of metallic landmine components and clutter for a broadband range of frequencies (kHz to tens of kHz). Results show the reconstructed spectral target signatures are very distinctive and hence potentially offer an efficient physical approach for landmine identification. The accuracy of the evaluated spectra is similarly verified using a uniform field forming sensor. © 2015 IOP Publishing Ltd.

https://iopscience.iop.org/article/10.1088/0957-0233/26/11/115501/pdf

Towards metal detection and identification for humanitarian demining using magnetic polarizability tensor spectroscopy

In this paper, the Fuzzy ARTMAP (FAM) neural network is used to classify metal detector signals into different categories for automated target discrimination. Feature extraction of the metal detector signals is conducted using a wavelet transform technique. The FAM neural network is then employed to classify the extracted features into different target groups. A series of experiments using individual FAM networks and a voting FAM network is conducted. Promising classification accuracy rates are obtained from using individual and voting FAM networks, respectively. The experimental outcomes positively demonstrate the effectiveness of the generated features, and of the FAM network in classifying metal detector signals for automated target discrimination tasks.

Feature extraction and classification of metal detector signals using the wavelet transform and the fuzzy ARTMAP neural network

In this paper, a methodology for target discrimination utilizing wavelet-based and morphological feature extraction is proposed. The proposed methodology is implemented into a landmine classification decision system utilizing metal detector array data as input. The classification performances of a number of feature vectors composed of different combinations of feature elements are assessed. This is conducted using a Fuzzy ARTMAP neural network classifier and majority voting decision fusion. The classification classes trialled during processing are target type and burial depth, both combined and individually. The majority of the results achieve correct classification percentages of above 80% both prior to and after decision fusion, with generally higher accuracies and lower misclassification percentages achieved after decision fusion.

A Target Discrimination Methodology Utilizing Wavelet-Based and Morphological Feature Extraction With Metal Detector Array Data

Landmines pose a significant problem in many countries around the world. Although technological systems such as metal detectors have been employed to combat these threats, many of these still require significant human interaction especially in the area of target and clutter discrimination. The aim of this research is to develop an automated decision making system for landmine detection. The initial stages of the research involves comparing various techniques for feature extraction to determine which methods provide the best representation for metal detector data to achieve improved target discrimination from background noise. This paper will focus on evaluating a technique utilizing the Continuous Wavelet Transform with false alarm rate and probability of detection used as performance measures.

Evaluation of the Continuous Wavelet Transform for Feature Extraction of Metal Detector Signals in Automated Target Detection

An automated decision system for landmine detection and discrimination is implemented and evaluated using metal detector array data. The techniques utilised include: a gradient based peak isolation method, wavelet transforms, fuzzy ARTMAP neural networks, and the generic majority voting scheme. The features selected for representing the input data are composed of the morphological and wavelet based features of the target signature responses. Classification experiments are conducted in an attempt to discriminate target type and burial depth according to two different methodologies. The results obtained are promising, with the implemented decision system achieving high probabilities of detection with reasonable false alarm rates, and exceptional discrimination before and after decision fusion with relatively low classification errors.

An Automated Decision System for Landmine Detection and Classification Using Metal Detector Signals

Ground Penetrating Radar (GPR) is considered as one of the promising technologies to address the challenges of detecting buried threat objects, particularly in military applications. The aim of this research is to design a potential solution to the task of target detection and classification using GPR. This paper focuses on the first stage of this task which is target detection. Three approaches for automated target detection are presented; a probabilistic approach, an artificial neural network with direct data input, and an artificial neural network with frequency spaced features. These techniques are applied to a preliminary data set with promising results.

/pdf/detection-of-targets-in-characteristic-gpr-sensor-data-using-4s6rbf8jlt.pdf

Minh Dao-Johnson Tran

Papers

Feature extraction and classification of metal detector signals using the wavelet transform and the fuzzy ARTMAP neural network

A Target Discrimination Methodology Utilizing Wavelet-Based and Morphological Feature Extraction With Metal Detector Array Data

Evaluation of the Continuous Wavelet Transform for Feature Extraction of Metal Detector Signals in Automated Target Detection

An Automated Decision System for Landmine Detection and Classification Using Metal Detector Signals

Detection of Targets in Characteristic GPR Sensor Data Using Machine Learning Techniques