The classification of skateboarding tricks via transfer learning pipelines

This study aims at classifying flat ground tricks, namely Ollie, Kickflip, Shove-it, Nollie and Frontside 180, through the identification of significant input image transformation on different transfer learning models with optimized Support Vector Machine (SVM) classifier. A total of six amateur skateboarders (20 ± 7 years of age with at least 5.0 years of experience) executed five tricks for each type of trick repeatedly on a customized ORY skateboard (IMU sensor fused) on a cemented ground. From the IMU data, a total of six raw signals extracted. A total of two input image type, namely raw data (RAW) and Continous Wavelet Transform (CWT), as well as six transfer learning models from three different families along with grid-searched optimized SVM, were investigated towards its efficacy in classifying the skateboarding tricks. It was shown from the study that RAW and CWT input images on MobileNet, MobileNetV2 and ResNet101 transfer learning models demonstrated the best test accuracy at 100% on the test dataset. Nonetheless, by evaluating the computational time amongst the best models, it was established that the CWT-MobileNet-Optimized SVM pipeline was found to be the best. It could be concluded that the proposed method is able to facilitate the judges as well as coaches in identifying skateboarding tricks execution.

A real-time approach of diagnosing rice leaf disease using deep learning-based faster R-CNN framework

The rice leaves related diseases often pose threats to the sustainable production of rice affecting many farmers around the world. Early diagnosis and appropriate remedy of the rice leaf infection is crucial in facilitating healthy growth of the rice plants to ensure adequate supply and food security to the rapidly increasing population. Therefore, machine-driven disease diagnosis systems could mitigate the limitations of the conventional methods for leaf disease diagnosis techniques that is often time-consuming, inaccurate, and expensive. Nowadays, computer-assisted rice leaf disease diagnosis systems are becoming very popular. However, several limitations ranging from strong image backgrounds, vague symptoms' edge, dissimilarity in the image capturing weather, lack of real field rice leaf image data, variation in symptoms from the same infection, multiple infections producing similar symptoms, and lack of efficient real-time system mar the efficacy of the system and its usage. To mitigate the aforesaid problems, a faster region-based convolutional neural network (Faster R-CNN) was employed for the real-time detection of rice leaf diseases in the present research. The Faster R-CNN algorithm introduces advanced RPN architecture that addresses the object location very precisely to generate candidate regions. The robustness of the Faster R-CNN model is enhanced by training the model with publicly available online and own real-field rice leaf datasets. The proposed deep-learning-based approach was observed to be effective in the automatic diagnosis of three discriminative rice leaf diseases including rice blast, brown spot, and hispa with an accuracy of 98.09%, 98.85%, and 99.17% respectively. Moreover, the model was able to identify a healthy rice leaf with an accuracy of 99.25%. The results obtained herein demonstrated that the Faster R-CNN model offers a high-performing rice leaf infection identification system that could diagnose the most common rice diseases more precisely in real-time.

The classification of EEG-based winking signals: a transfer learning and random forest pipeline

Brain Computer-Interface (BCI) technology plays a considerable role in the control of rehabilitation or peripheral devices for stroke patients. This is particularly due to their inability to control such devices from their inherent physical limitations after such an attack. More often than not, the control of such devices exploits electroencephalogram (EEG) signals. Nonetheless, it is worth noting that the extraction of the features and the classification of the signals is non-trivial for a successful BCI system. The use of Transfer Learning (TL) has been demonstrated to be a powerful tool in the extraction of essential features. However, the employment of such a method towards BCI applications, particularly in regard to EEG signals, are somewhat limited. The present study aims to evaluate the effectiveness of different TL models in extracting features for the classification of wink-based EEG signals. The extracted features are classified by means of fine-tuned Random Forest (RF) classifier. The raw EEG signals are transformed into a scalogram image via Continuous Wavelet Transform (CWT) before it was fed into the TL models, namely InceptionV3, Inception ResNetV2, Xception and MobileNet. The dataset was divided into training, validation, and test datasets, respectively, via a stratified ratio of 60:20:20. The hyperparameters of the RF models were optimised through the grid search approach, in which the five-fold cross-validation technique was adopted. The optimised RF classifier performance was compared with the conventional TL-based CNN classifier performance. It was demonstrated from the study that the best TL model identified is the Inception ResNetV2 along with an optimised RF pipeline, as it was able to yield a classification accuracy of 100% on both the training and validation dataset. Therefore, it could be established from the study that a comparable classification efficacy is attainable via the Inception ResNetV2 with an optimised RF pipeline. It is envisaged that the implementation of the proposed architecture to a BCI system would potentially facilitate post-stroke patients to lead a better life quality.

The classification of motor imagery response: an accuracy enhancement through the ensemble of random subspace k-NN

Brain-computer interface (BCI) is a viable alternative communication strategy for patients of neurological disorders as it facilitates the translation of human intent into device commands. The performance of BCIs primarily depends on the efficacy of the feature extraction and feature selection techniques, as well as the classification algorithms employed. More often than not, high dimensional feature set contains redundant features that may degrade a given classifier's performance. In the present investigation, an ensemble learning-based classification algorithm, namely random subspace k-nearest neighbour (k-NN) has been proposed to classify the motor imagery (MI) data. The common spatial pattern (CSP) has been applied to extract the features from the MI response, and the effectiveness of random forest (RF)-based feature selection algorithm has also been investigated. In order to evaluate the efficacy of the proposed method, an experimental study has been implemented using four publicly available MI dataset (BCI Competition III dataset 1 (data-1), dataset IIIA (data-2), dataset IVA (data-3) and BCI Competition IV dataset II (data-4)). It was shown that the ensemble-based random subspace k-NN approach achieved the superior classification accuracy (CA) of 99.21%, 93.19%, 93.57% and 90.32% for data-1, data-2, data-3 and data-4, respectively against other models evaluated, namely linear discriminant analysis, support vector machine, random forest, Naïve Bayes and the conventional k-NN. In comparison with other classification approaches reported in the recent studies, the proposed method enhanced the accuracy by 2.09% for data-1, 1.29% for data-2, 4.95% for data-3 and 5.71% for data-4, respectively. Moreover, it is worth highlighting that the RF feature selection technique employed in the present study was able to significantly reduce the feature dimension without compromising the overall CA. The outcome from the present study implies that the proposed method may significantly enhance the accuracy of MI data classification.

The classification of movement intention through machine learning models: the identification of significant time-domain EMG features

Electromyography (EMG) signal is one of the extensively utilised biological signals for predicting human motor intention, which is an essential element in human-robot collaboration platforms. Studies on motion intention prediction from EMG signals have often been concentrated on either classification and regression models of muscle activity. In this study, we leverage the information from the EMG signals, to detect the subject's intentions in generating motion commands for a robot-assisted upper limb rehabilitation platform. The EMG signals are recorded from ten healthy subjects' biceps muscle, and the movements of the upper limb evaluated are voluntary elbow flexion and extension along the sagittal plane. The signals are filtered through a fifth-order Butterworth filter. A number of features were extracted from the filtered signals namely waveform length (WL), mean absolute value (MAV), root mean square (RMS), standard deviation (SD), minimum (MIN) and maximum (MAX). Several different classifiers viz. Linear Discriminant Analysis (LDA), Logistic Regression (LR), Decision Tree (DT), Support Vector Machine (SVM) and k-Nearest Neighbour (k-NN) were investigated on its efficacy to accurately classify the pre-intention and intention classes based on the significant features identified (MIN and MAX) via Extremely Randomised Tree feature selection technique. It was observed from the present investigation that the DT classifier yielded an excellent classification with a classification accuracy of 100%, 99% and 99% on training, testing and validation dataset, respectively based on the identified features. The findings of the present investigation are non-trivial towards facilitating the rehabilitation phase of patients based on their actual capability and hence, would eventually yield a more active participation from them.

A review on non-invasive hypertension monitoring system by using photoplethysmography method

Hypertension, the abnormal elevation of blood pressure, is one of the chronic disease that usually comes with no symptom and signal. As the systolic blood pressure (SBP) over 140 mm Hg or diastolic blood pressure (DBP) is over 90 mmHg, it is considered as hypertension. The purpose of this paper is to determine a method to early diagnose of hypertension by monitoring the SBP, DBP, and heart rate (HR) non-invasively. Although accurate measurement of BP and HR of a person can be obtained invasively, the measuring probe needs to place under patient’s skin and it would cause wound. Therefore, this paper review on methods to measure BP and HR non-invasively. External pressures are needed to induce to the artery in order to measure BP and HR by using auscultatory and oscillometric methods, hence, a pressure cuff is used for measuring BP. The pressure cuff will restrict the motion of patient and it is not suitable for continuous monitoring. Pulse transit time (PTT) and photoplethysmography (PPG) methods are introduced to measuring BP non-invasively without cuff. The limitation of PTT over PPG is PTT needs both PPG waveform and ECG waveform to estimate BP, and artificial phase lag is occurred which will affect the reliability of the measured result. Therefore, for long term monitoring hypertension, non-invasively, by using photoplephymosgraphy method is preferred since it enables continuous monitoring without cuff and it is only one waveform, which is PPG waveform, is needed to estimate the BP as well as HR. 

The Development of the Putt.It.In Monitoring Device and the Establishment of Its Reliability: A Solution for Putting-In Analysis in Golf

Background: The accurate transfer of information on the athletes’ performance in any sport is essential in enhancing the performance and overall coaching process. The provision of such information is favourable only if it is reliable. A cost-effective golf putting monitoring device namely the Putt.It.In was developed for analysing a golfers’ putting performance. Objectives: This study aims to investigate the reliability of the device in measuring the backswing distance, front swing distance, clubhead speed, ideal front swing distance and swing angle. Methods: A semi- professional golfer (30 years of age ± 5.0 years’ experience) executed four strokes repeatedly from a distance of 2 m and 1 m using a Ram Zebra Mallet putter on a PGM golf mat. Kolmogorov/Smirnov test was utilised to ascertain the reliability of the application in measuring the aforementioned parameters over test re-test between first two strokes of 2 m distance and the last two strokes of 1 m distance. Results: The Kolmogorov/Smirnov test re-test suggests that there is no significant difference between first two meters strokes p > 0.05, and second 1-meter strokes p > 0.05 highlighting its ability to recognise the pattern of the strokes applied in the four successive strokes. Conclusion: The Putt.It.In monitoring device is found to be reliable in measuring the backswing distance, front swing distance, clubhead speed, ideal front swing distance and swing angle. Professional and semi-professional golfers as well coaches could consider Putt.It.In device in monitoring strokes related parameters to enhance their performance due to its effectiveness in providing information on putting performance.