Search

Weeds affects crops health as it shares water and nutrients from the soil, as a result it decreases crop yield. Manual weedicide spray through bag-pack is hazardous to human health. Localized autonomous weedicide spray through aerial spraying units can help save water, weedicide chemical and effect less on human health. Such systems require multi-spectral cues to classify crop, weed, and soil surface. Our focus in this paper is on the detection of weeds in the sugar beet crop, using air-borne multispectral camera sensors, which is considered as an alternative crop to sugarcane to obtain sugar in Pakistan. We developed a new framework for weed identification; a patch-based classification approach as appose to semantic segmentation that is more realistic for real-time intelligent aerial spraying systems. Our approach converts 3-class pixel classification problem into a 2-class crop-weed patch classification problem which in turns improves crop and weed classification accuracy. For classification, we developed a new VGG-Beet convolutional neural network (CNN), which is based on generic VGG16 (visual graphics group) CNN model with 11 convolutional layers. For experiments, we captured a sugar beet dataset with 3-channel multispectral sensor with a ground sampling distance (GSD) of 0.2 cm/pixel and a height of 4 meters. For better comparison, we used two publicly available sugar beet crop aerial imagery datasets, captured using a 5-channel multispectral sensor and a 4-Channel multispectral sensor with a ground sampling distance of 1cm and a height of 10 meters. We observed that patch-based method is more robust to different lighting conditions. To produce low cost weed detection system usage of Agrocam sensor is recommended, for higher accuracy Red Edge and Sequoia multispectral sensors with more channels should be deployed. We observed higher crop-weed accuracy and lower testing time for our patch-based approach as compared to U-Net and Deeplab based semantic segmentation networks.

Experimentation and analysis in brain-computer interface (BCI) has increasingly been receiving quite some consideration as a substitute communication possibility for patients who are severely paralyzed in the last few years. To measure brain activities using optical signals a fairly new and non-invasive brain imaging tool can be put to test know as Functional near-infrared spectroscopy (fNIRS). Comparability low cost, safety, portability and wear ability are some of the main advantages of imaging of brain using this non-invasive modality. We propose in this paper to apply this relatively new non-invasive fNIRS technique to make an image of brain activities during four different mental tasks. These tasks include Mental Arithmetic (MA), Motor Imagery (i.e. Left-Hand and Right-Hand Motor Imagery) and Rest. fNIRS data used is an open access dataset which was collected by Continuous-wave imaging system (NIR Scout NIRx GmbH, Berlin, Germany) with the sampling frequency of 10 Hz. The research we have done in this paper Data synchronization is performed before the data is preprocessed. After the preprocessing and signal analysis of data our results shows hemodynamic behavior of multiple patterns during the tasks performed. These unique patterns of hemodynamic behavior can be used to differentiate and distinguish different task. We were able to compare, differentiate and distinguish the brain signal activities captured while performing 4 different tasks using 3 different classifiers i.e. Linear Discriminant Analysis (LDA), Support Vector Machine (SVM) and K Nearest Neighbor (KNN). The average classification accuracy of above 90% is achieved by using K Nearest Neighbors (KNN).

Analysis of visual cues for fruit classification and sorting allows to automate the visual inspection and packaging process in agricultural applications that is performed so far by human workers. Challenges for automated multi-class sorting systems are similarity in color and shape of different fruit varieties and variation among the same category of fruit. A major constraint in using well known deep neural networks for fruit classification arises because deep neural networks require large training datasets for achieving high accuracies which are generally not available in case of agricultural products especially various fruits and vegetable varieties. A thorough analysis is required to find an appropriate combination of various handcrafted features that could give precise and accurate classification results for small datasets. This paper investigates the use of various handcrafted visual features for fruit classification using traditional machine learning techniques. Different color, shape and texture features are analyzed by comparing the results obtained from six supervised machine learning techniques including K nearest neighbors, Support Vector Machines, Naive Bayes, Linear Discriminant Analysis, Decision Trees and Feed forward back propagation neural network. We propose a novel combination of Hue, Color-SIFT, Discrete Wavelet Transform and Haralick features in fruit classification problem that outperforms other handcrafted visual features. This feature combination is found to be invariant to rotation and illumination effects and works well with intra class variations providing good results for identifying subcategories of fruits along with high classification accuracies obtained for difficult fruit categories that are visually similar. It is found that Color-SIFT features alone work very well for fruit classification problem by outperforming other individual handcrafted features. Our approach is trained and tested on publicly available Fruits 360 dataset. Out of different classifiers best results are obtained using Back Propagation Neural Network, SVM and KNN classifier with classification accuracies between 99% and 100%.

The study proposed the classification and recognition of hand gestures using electromyography (EMG) signals for controlling the upper limb prosthesis. In this research, the EMG signals were measured through an embedded system by wearing a band of MYO gesture control. In order to observe the behavior of these change movements, the EMG data was acquired from 10 healthy subjects (five male and five females) performing four upper limb movements. After extracting EMG data from MYO, the supervised classification approach was applied to recognize the different hand movements. The classification was performed with a 5-fold cross-validation technique under the supervision of Quadratic discriminant analysis (QDA), support vector machine (SVM), random forest, gradient boosted, ensemble (bagged tree), and ensemble (subspace K-Nearest Neighbors) classifier. The execution of these classifiers shows the overall accuracy of 83.9% in the case of ensemble (bagged tree) which is higher than other classifiers. Additionally, in this research an embedded system-based classification approach of hand movement was used for designing an upper limb prosthesis. This approach is different than previous techniques as MYO is used with an external Bluetooth module and different libraries that make its movement and performance boundless. The results of this study also inferred the operations which were easy for hand recognition and can be used for developing a powerful, efficient, and flexible prosthetic design in the future.

Foot movement during walking, jogging, running, jumping, skidding, and climbing etc. is an important emphasis in lower limb prosthesis. This work presents the estimation of gait phases of walking persons on flat terrain. Different phases of swing and stance of foot gait cycle have been considered with sensing unit consisting of gyroscope and potentiometer. Arduino microcontroller has been incorporated to process the values received from the sensing unit using rule based algorithm. Sensor values are processed using ‘Extra Putty’ software and the results are observed at serial monitor. This project has been tested on a person and real time gait phases have been drawn. Gait phases of forward and reverse walk have been successfully identified. The work is useful to identify the balanced/unbalanced state of persons during normal walk.

This paper presents a human-computer interface system using Electromyography (EMG) signals to design a computer-based video game that enhances the muscle strength and can be used for rehabilitation of fist, wrist or forearm muscles. This game encompasses the rehabilitation of the muscle in itself by asking the user to move the muscle and measure the potential difference generated due to action potential using MyoWare muscle sensors kit. This newly generated raw EMG signal is passed on to Arduino, which is interfaced with a computer through a Graphical User Interface (GUI) based computer game, coded and designed in MATLAB. The gaming console presents an interesting and enjoyable platform as a modern treatment approach for stroke rehabilitation and provides an opportunity to practice activities that are not or cannot be practiced within the clinical environment. It encourages a higher number of repetitions than traditional therapy tasks without the supervision of an expert. However, due to the higher computational delay of 2 to 5 seconds between two consecutive input operations, MATLAB is not suitable for creating a real-time virtual gaming environment. The primary objective of the paper is to design and evaluate the effects of an interactive video game environment for upper limb muscle rehabilitation. Secondary outcomes included process limitations and adverse events based on the current equipment and techniques used.

Mimicking ankle angles continuously and accurately during gait cycle for lower limb prosthesis is of prime importance in prosthesis control. A robust control system is thus desired to mimic natural gait cycle of a healthy human for an amputee. This research focuses on simulating a control system which mimics natural ankle angles of human ankle biomechanics. The controller mimics a reference ankle angle input as output with maximum accuracy for lower limb prosthesis developed earlier. This control system can be further advanced to mimic human ankle torque and power profiles as well.

Diabetic Retinopathy (DR) is an eye disorder that progressively leads to vision loss due to high glucose causing impairment of retinal blood vessels (BVs). ‘Retinal Bright Lesions’ such as ‘Hard Exudates’ (HEs) are plasma leakages from rapture retinal capillaries. HEs appear as hard, waxy, yellowish deposits from tiny spots to fat patches and signify moderate-severe Non-Proliferative Diabetic Retinopathy (NPDR). This paper proposes a simple, compact and computationally inexpensive technique for detection and classification of HEs using Digital Image Processing Techniques on digital fundus images complements and Artificial Neural Networks (ANN). The proposed technique unfolds through five stages i.e. Pre-processing, coarse detection, optimization, features detection & extraction followed by classification. ‘Speed Up Robust Features’ (SURF) algorithm has been used for features detection & extraction while ‘Feed-Forward Back-propagation’ (FFBP) ANN has been used for classification. The proposed technique has yielded 98.7% ‘Sensitivity’ (SE), 97.5% ‘Specificity’ (SP) and 97.7% ‘Accuracy’ (AC) on ‘DIARETDB1’ fundus images.

Artificial immune systems (AIS) are intelligent algorithms derived from the principles inspired by the human immune system. In this study, electroencephalography (EEG) signals for four distinct motor movements of human limbs are detected and classified using a negative selection classification algorithm (NSCA). For this study, a widely studied open source EEG signal database (BCI IV-Graz dataset 2a, comprising nine subjects) has been used. Mel frequency cepstral coefficients (MFCCs) are extracted as selected features from recorded EEG signals. Dimensionality reduction of data is carried out by applying two hidden layered stacked auto-encoder. Genetic algorithm (GA) optimized detectors (artificial lymphocytes) are trained using negative selection algorithm (NSA) for detection and classification of four motor movements. The trained detectors consist of four sets of detectors, each set is trained for detection and classification of one of the four movements from the other three movements. The optimized radius of detector is small enough not to mis-detect the sample. Euclidean distance of each detector with every training dataset sample is taken and compared with the optimized radius of the detector as a nonself detector. Our proposed approach achieved a mean classification accuracy of 86.39% for limb movements over nine subjects with a maximum individual subject classification accuracy of 97.5% for subject number eight.

Experimentation and analysis of Functional near-infrared spectroscopy (fNIRS) in Brain-Computer Interface (BCI) has increasingly been studied as a communication possibility for patients who are severely paralyzed. This study has applied this technique to distinguish brain activities during four different mental tasks. These tasks include Mental Arithmetic (MA), Motor Imagery of Left-Hand (LHMI) and Right-Hand (RHMI) and Rest. fNIRS data used is from an open access dataset of 29 individuals which was collected by Continuous-wave imaging system (NIR Scout). In this research Data integration is performed before the data is preprocessed. Usual preprocessing is done using Butterworth filter to minimize or eliminate any unwanted signal distortion. After that an extensive signal analysis is done in which six different statistical features (Signal Mean (SM), Skewness (SK), Kurtosis (KR), Standard Deviation (SD), Signal Peak (SP), and Signal Variance (SV)) are obtained in the time domain and 13 Mel Frequency Cepstral Coefficients (MFCC) features are obtained from the frequency domain. As per literature review, MFCC has never been used as feature towards classification of fNIRS signal, which is a novel contribution towards this study. Separate Classification analysis is performed on each domain features. We were able to compare, differentiate and distinguish the brain signal activities captured while performing four different tasks using three different classifiers i.e. Linear Discriminant Analysis (LDA), Support Vector Machine (SVM) and K Nearest Neighbor (KNN). The average classification accuracy of 90.54% is achieved from K Nearest Neighbors (KNN) using the time domain features and accuracy achieved from Support Vector Machine (SVM) using the frequency domain features is 95.7%. Comparison with benchmark study shows the efficiency of MFCC as suitable features for improved classification accuracy.

this paper proposes a scheme for conditioning and processing of signal from a pre amplifier EMG sensor. An electronic system is devised to help the experimenter working with EMG sensor by Motion Labs. Signal conditioning and processing strategy is carried out in three main stages; first one being the amplification of signal acquired from a muscle activity. In the next stage, this amplified signal is rectified and in the end, it is further processed with a controller. The developed system would be helpful for both, the candidates interested in research and experimentation of signal conditioning and processing of EMG sensor manufactured by Motion Labs. The developed controller algorithm is used to convert frequency based sensor input into magnitude based output which is then filtered to smoothen the response.

Feature extraction is a pronounced method to infer the information utility which is concealed in electromyography (EMG) signal to study the characteristic properties and behavior of signal. This study gives a comparative analysis of thirteen complete and most up-to-date EMG feature signals in Time-domain and Frequency-domain. Particularly, the EMG signals are obtained from a device MYO gesture control on an embedded system. For this purpose, four healthy male volunteers are considered to perform four different hand movements based on stationary, double tap, single finger movement and finger spread. To be a successful classification of these EMG features in both domains, we prefer attribute selected classifier as it gives the better performance and higher rate of accuracy i.e. 93.8%. The experimental results prove that features in time-domain are superfluity and redundant while features in frequency-domain (measured by statistical parameters of EMG power spectral density) show the ultimate dominance and signal characterization. The findings of this study are highly beneficial for further use in order to predict the behavior of EMG in pattern recognition and in classification of EMG signals for assistive devices or in powered human arm prosthetics.

Brain Computer Interface (BCI) determines the intent of the user from a variety of electrophysiological signals. These signals, Slow Cortical Potentials, are recorded from scalp, and cortical neuronal activity is recorded by implanted electrodes. This paper is focused on design of an embedded system that is used to control the finger movements of an upper limb prosthesis using Electroencephalogram (EEG) signals. This is a follow-up of our previous research which explored the best method to classify three movements of fingers (thumb movement, index finger movement, and first movement). Two-stage logistic regression classifier exhibited the highest classification accuracy while Power Spectral Density (PSD) was used as a feature of the filtered signal. The EEG signal data set was recorded using a 14-channel electrode headset (a noninvasive BCI system) from right-handed, neurologically intact volunteers. Mu (commonly known as alpha waves) and Beta Rhythms (8–30 Hz) containing most of the movement data were retained through filtering using “Arduino Uno” microcontroller followed by 2-stage logistic regression to obtain a mean classification accuracy of 70%.

Background. Brain computer interface (BCI) is a combination of software and hardware communication protocols that allow brain to control external devices. Main purpose of BCI controlled external devices is to provide communication medium for disabled persons. Now these devices are considered as a new way to rehabilitate patients with impunities. There are certain potentials present in electroencephalogram (EEG) that correspond to specific event. Main issue is to detect such event related potentials online in such a low signal to noise ratio (SNR). In this paper we propose a method that will facilitate the concept of online processing by providing an efficient filtering implementation in a hardware friendly environment by switching to finite impulse response (FIR). Main focus of this research is to minimize latency and computational delay of preprocessing related to any BCI application. Four different finite impulse response (FIR) implementations along with large Laplacian filter are implemented in Xilinx System Generator. Efficiency of 25% is achieved in terms of reduced number of coefficients and multiplications which in turn reduce computational delays accordingly.

This paper presents the design of a stair climbing fixed flipper unmanned ground vehicle (UGV) for urban search and rescue purposes. Mobile flippers are being used in certain UGVs for enhanced mobility in rough terrains, however, the control algorithm of these platforms is complex. To add this enhanced mobility in the UGV and to reduce the intricacy of the control algorithm, anterior end of the tracks are lifted up which enables the UGV to pass over obstacles with relative ease. To prevent the rollover of UGV while moving on an inclined surface, an image processing algorithm was developed which halts the motion of UGV if the calculated slope exceeds the threshold value with a maximum error of about 8%. Furthermore, left and right track velocities along with the turn radius were also calculated.

This paper presents an application of a novel approach for detecting and tracking an object with a 2 DOF robotic manipulator which can be equipped with an array of electrically controlled actuators. The said approach utilizes the Image Based Visual Servoing (IBVS) technique. The developed system is able to determine the object pose in real time from features in the image. Object is detected using shaped based approach algorithms of image processing. The position and orientation of the world coordinates of the object being tracked are calculated from the coordinates of the object in image plane using camera's intrinsic and extrinsic parameters. Experimental results demonstrate the effectiveness of this proposed approach.

Recent advancements in brain computer-interfacing (BCI) and neuro-robotics have played an indispensable role for people suffering from neural injuries to expect better quality of life by restoring sensory functions and replacement of neuro-muscular pathways as BCI systems work on imagination of movements to control prosthetic limbs. In this research, multiple combinations of features and classifiers have been used to classify electroencephalographic (EEG) signals in order to acquire maximum classification accuracy for a four class EEG based BCI system. Knee and ankle joint movements are executed in a predefined manner during which EEG signals are acquired from sensorimotor cortex of four healthy test subjects, using Emotiv headset. After removing artifacts, five features; average band power, peak, kurtosis, mean and skewness are calculated. Multiple combinations of calculated features are used to classify ankle dorsi-planter flexion and knee extension-flexion movements using five different classifiers; Naive Bayes, A-nearest neighbor (ANN), linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), and support vector machine (SVM). It was found that features set comprising; average band power, peak and mean value yielded significantly higher classification accuracy of 81.31% among all other combinations whereas in classifiers, maximum accuracy of 81.48% is obtained through LDA as compared to other classifiers. These novel findings clearly exhibit the feasibility of achieving good classification accuracies using average band power, mean and peak values of EEG signals as feature vectors and LDA as classifier.

With the evolution of broadband mobile networks towards LTE and beyond, the support for the internet and internet based services is growing. However, the size and operational costs of mobile networks are also growing. Self Organizing Networks (SON) are introduced as a part of the specifications of the LTE standard with the purpose of reducing the Operation and Maintenance costs of the mobile networks. This paper introduces a novel framework for automated Radio Resource Management (RRM) in LTE SON. This framework deals with the self-optimization and self-healing features of SON. The data mining technique of linear regression has been used to derive the functional relationship, known as model, between Key Performance Indicators and RRM parameters. The proposed framework uses this model in two ways: first, for network monitoring, which is the first step of the self-healing procedure and secondly, to devise a handover auto-tuning algorithm as part of the self-optimization procedure. The detailed results obtained for the finished case studies, demonstrate the effectiveness and usefulness of this approach.

Recent developments in Brain Computer Interfacing (BCI) and neuroprosthetics have played a vital role for disable people to expect better life quality. In this contribution Electroencephalographic (EEG) signals acquired from six healthy test subjects, are used for the offline analysis of BCI through classification of four lower limb movements including talocrural (ankle) joint dorsi-planter flexion and knee joint extension-flexion. Fourteen channel Emotive EPOC head set is used to acquire EEG signals from sensorimotor cortex area of brain, using a particular data acquisition timeline protocol. Features are extracted in time domain from raw EEG data. Power spectral density, variance, mean value and kurtosis features are applied on raw EEG signals. Multiple classification algorithms are implemented for discrimination of four lower limb movements within data set. The paper uses Quadratic discriminant analysis, Naive Bayes and Support vector machine classifiers to stratify the movement intent of lower limb. Maximum classification accuracies achieved through various classifiers are; 86.35% with average band power & QDA, 84.38% with mean value & QDA and 78.13% with power spectral density & Quadratic-SVM. The presented findings are optimistic in making the path easier towards the development of BCIs with rich EEG based control signals using noninvasive technology.

A type of unmanned ground vehicle (UGV) called Versatile Terrain Mobility Platform (VTAMP) has been designed with the idea of providing a dynamic mobile platform with the unique ability of traversing through rough terrains, and also to pass over obstacles otherwise proving to be a difficult task for most tele-operated ground vehicles. Increased maneuverability is achieved by using active arm-like tracks called “ flippers”. By being able to operate each flipper independently, we can operate VTAMP in different modes with the operator control unit (OCU) which in our scenario is a laptop. This is achieved through feedback loop on each flipper using encoders. The ability to c hange configuration facilitates VTAMP to be able to adapt to all kinds of terrains. The basic architecture of VTAMP is presented first, followed by its different modes of operation by possible flipper arrangements and its control scheme. An algorithm was made to prevent accidental collision between flippers. In addition to this, the turning radius of the VTAMP was also calculated according to the track speeds of both sides which were then verified through experiment.

In recent times there is a prevalence in amputations in the developing world due to lack of proper medical treatment of diseases & injuries sustained in natural disasters or war; creating a need for an optimal solution that is technically sound and is affordable. This paper discusses the available solutions in the market and then proposes a new design for prosthetic hands, incorporating the morphological features of the actual human hand. Each finger, actuated by a slider mechanism, has a single degree of freedom. While the thumb design utilizes worm gear driven mechanisms to feature a double degree of freedom. The morphology of the developed device is that of the adult male human right hand. The motors for actuation are installed inside the palm of the hand while the overall weight of the prosthetic device is 453 g. The control is generated by the forward kinematics model assessing the relation between the slider position and the rotation angles of the metacarpophalangeal (MCP) joints of the individual fingers. A feedback control loop provides the adaptive grip. The developed prosthesis design can be used to interface with either an EMG or EEG-based control scheme.

This paper presents our approach to staircase detection and perception using monoscopic vision. The simulation of the proposed algorithm is conducted on a 4 degree of freedom teleoperated vehicle. The presented solution can detect a staircase and calculate its angle of inclination thus aiding the operator in setting up controls for the vehicle. The approach relies on multiple 5×5 masks to actively predict staircase from the front then by a combination of blob detecting algorithms predict the angle of inclination. One of the challenges was to make the algorithm efficient so that it can be implemented on low power embedded systems. Results of simulations on various acquired images have been discussed within the contents of the paper. The results have been validated by monitoring values over multiple & different staircases. MATLAB's image processing toolbox has been used to conduct simulations of the algorithm.

Electroencephalographic (EEG) signals from test subjects are used for BCI analysis and classification of the four upper limb movements. Brain computer interface (BCI) system provides better neural control to the user, over two antebrachium and two carpus movement. This paper uses Epoc Emotiv head set with fourteen electrodes to acquire signal from the motor area of scalp. A particular protocol for data acquisition is followed. Our method is based on the time domain feature extraction techniques. Activity feature is applied on the signals. In order to discriminate between four upper limb movements within the data set, the paper uses Quadratic support vector machine, RBF SVM and Multilayer perceptron. The best classification accuracy of two antebrachium and two carpus movement is achieved by Quadratic SVM and RBF SVM when Gaussian window is used to calculate the activity feature vector. The number of correctly classified instants in term of percentage is 76.92% and 75.96% for quadratic SVM and RBF SVM. Current classifiers show promising results, where as to make conclusion more generalized and to enhance the classification accuracy, further experiments need to be carried out.