Parkinson’s Disease (PD) is a complex neurodegenerative disorder characterized by a spectrum of motor and non-motor symptoms, prominently featuring the freezing of gait (FOG), which significantly ...impairs patients’ quality of life. Despite extensive research, the precise mechanisms underlying FOG remain elusive, posing challenges for effective management and treatment. This paper presents a comprehensive meta-analysis of FOG prediction and detection methodologies, with a focus on the integration of wearable sensor technology and machine learning (ML) approaches. Through an exhaustive review of the literature, this study identifies key trends, datasets, preprocessing techniques, feature extraction methods, evaluation metrics, and comparative analyses between ML and non-ML approaches. The analysis also explores the utilization of cueing devices. The limited adoption of explainable AI (XAI) approaches in FOG prediction research represents a significant gap. Improving user acceptance and comprehension requires an understanding of the logic underlying algorithm predictions. Current FOG detection and prediction research has a number of limitations, which are identified in the discussion. These include issues with cueing devices, dataset constraints, ethical and privacy concerns, financial and accessibility restrictions, and the requirement for multidisciplinary collaboration. Future research avenues center on refining explainability, expanding and diversifying datasets, adhering to user requirements, and increasing detection and prediction accuracy. The findings contribute to advancing the understanding of FOG and offer valuable guidance for the development of more effective detection and prediction methodologies, ultimately benefiting individuals affected by PD.
Fencing in livestock management is essential for location and movement control yet with conventional methods to require close labour supervision, leading to increased costs and reduced flexibility. ...Consequently, virtual fencing systems (VF) have recently gained noticeable attention as an effective method for the maintenance and control of restricted areas for animals. Existing systems to control animal movement use audio followed by controversial electric shocks which are prohibited in various countries. Accordingly, the present work has investigated the sole application of audio signals in training and managing animal behaviour. Audio cues in the range of 125–17 kHz were used to prohibit the entrance of seven Hebridean ewes from a restricted area with a feed bowl. Two trials were performed over the period of a year which were video recorded. Sound signals were activated when the animal approached a feed bowl and a restricted area with no feed bowl present. Results from both trials demonstrated that white noise and sounds in the frequency ranges of 125–440 Hz to 10–17 kHz successfully discouraged animals from entering a specific area with an overall success rate of 89.88% (white noise: 92.28%, 10–14 kHz: 89.13%, 15–17 kHz: 88.48%, 125–440 Hz: 88.44%). The study demonstrated that unaided audio stimuli were effective at managing virtual fencing for sheep.
Display omitted
•Monitoring animals through human observation is time consuming and labour intensive.•Smart devices mounted on the animals, embedded with predictive models are needed.•Dataset ...diversity using two types of accelerometers attached to animals’ collars.•Convolutional Neural Networks to predict active, grazing, and inactive behaviours.•Use of Transfer learning to evaluate the generalization of the pre-trained model.
Machine learning and sensor devices lined up with agriculture for the development of systems can efficiently provide real-time knowledge on animal behavior without the need of intense human observation, which is time consuming and labor demanding. In this study, we propose an intelligent system to classify the three important activities of sheep namely, “grazing”, “active”, and “inactive” states. We acquire primary data from Hebridean ewes using two types of sensors to capture the required activities. To address the problem of sensor heterogeneity in data and sensor orientation placement, we use convolutional neural networks (CNN) in conjunction with hand-crafted features, to improve the model generalization, specifically in terms of sensor orientation and position. Additionally, we utilise transfer learning (TL) for the model generalisation, which indicates substantial potential in future studies concerning animal activity recognition. More specifically, we use the TL to enable the reusability of pre-trained model for purely unseen data without performing model training and data labelling, which are highly time-consuming tasks. We performed experiments on datasets using CNN with automatic learned features, and on datasets using additional hand-crafted features. Our method obtained an overall accuracy of 98.55% on the source data (i.e., dataset captured via first sensor), and 96.59% on the target data (i.e., dataset captured via second sensor) when using the datasets which comprised the supplementary features. On the other hand, when CNN was applied to the raw datasets without any additional features, an accuracy of 97.46% and 94.79% was obtained on the source and target dataset respectively. This study is the first of its kind to propose convolutional neural network-based TL for the sheep activity recognition and demonstrate the significance of proposed approach in the context of data capturing, data labelling, and heterogeneity of sensor devices.
•Freezing of Gait (FoG) is a motor symptom of Parkinson's disease.•It negatively impacts on the quality of life of people suffering from this disease.•This study focus on prediction of the onset of a ...FoG event using machine learning.•The effect of signal features and window size in FoG prediction is investigated.•Balanced classification is attained using RBF-SVM and a 3 s transition period.
Freezing of Gait (FoG) is a motor symptom of Parkinson's disease (PD) that frequently occurs in the long-term sufferers of the disease. FoG may result to nursing home admission as it can lead to falls, and therefore, it impacts negatively on the quality of life. The focus of this study is the systematic evaluation of machine learning techniques in conjunction with varying size time windows and time/frequency domain feature sets in predicting a FoG event before its onset. In the experiments, the Daphnet FoG dataset is used to benchmark performance. This consists of accelerometer signals obtained from sensors mounted on the ankle, thigh and trunk of the PD patients. The dataset is annotated with instances of normal activity events, and FoG events. To predict the onset of FoG, the dataset is augmented with an additional class, termed ‘transition’, which relates to a manually defined period prior to the occurrence of a FoG episode. In this research, five machine learning models are used, namely, Random Forest, Extreme Gradient Boosting, Gradient Boosting, Support Vector Machines using Radial Basis Functions, and Neural Networks. Support Vector Machines with Radial Basis kernels provided the best performance achieving sensitivity values of 72.34%, 91.49%, 75.00%, and specificity values of 87.36%, 88.51% and 93.62%, for the FoG, transition and normal activity classes, respectively.
Animal activity recognition is an important topic that facilitates understanding of animal behavior that is useful for analyzing and classifying their wellbeing. Research studies have been reporting ...the use of animal activity as an effective indicator of their health state. This survey focuses on recent advancements in machine intelligence utilizing wearable devices for sheep activity recognition. We summarise existing works focusing on various types of sensors used in agricultural sheep activity recognition. Furthermore, data segmentation methods used in each study, followed by the potential recommendations on window size and sample rate selection are addressed in detail. Finally, we present the features being identified as significant along with an overview of machine learning algorithms used in the domain of sheep activity recognition using accelerometer data.
With the growth in world population, there is an increasing demand for food resources and better land utilisation, e.g., domesticated animals and land management, which in turn brought about ...developments in intelligent farming. Modern farms rely upon intelligent sensors and advanced software solutions, to optimally manage pasture and support animal welfare. A very significant aspect in domesticated animal farms is monitoring and understanding of animal activity, which provides vital insight into animal well-being and the environment they live in. Moreover, “virtual” fencing systems provide an alternative to managing farmland by replacing traditional boundaries.This thesis proposes novel solutions to animal activity recognition based on accelerometer data using machine learning strategies, and supports the development of virtual fencing systems via animal behaviour management using audio stimuli. The first contribution of this work is four datasets comprising accelerometer gait signals. The first dataset consisted of accelerometer and gyroscope measurements, which were obtained using a Samsung smartphone on seven animals. Next, a dataset of accelerometer measurements was collected using the MetamotionR device on 8 Hebridean ewes. Finally, two datasets of nine Hebridean ewes were collected from two sensors (MetamotionR and Raspberry Pi) comprising of accelerometer signals describing active, inactive and grazing activity of the animal. These datasets will be made publicly available as there is limited availability of such datasets. In respect to activity recognition, a systematic study of the experimental setup, associated signal features and machine learning methods was performed. It was found that Random Forest using accelerometer measurements and a sample rate of 12.5Hz with a sliding window of 5 seconds provides an accuracy of above 96% when discriminating animal activity. The problem of sensor heterogeneity was addressed with transfer learning of Convolutional Neural Networks, which has been used for the first time in this problem, and resulted to an accuracy of 98.55%, and 96.59%, respectively, in the two experimental datasets. Next, the feasibility of using only audio stimuli in the context of a virtual fencing system was explored.Specifically, a systematic evaluation of the parameters of audio stimuli, e.g., frequency and duration, was performed on two sheep breeds, Hebridean and Greyface Dartmoor ewes, in the context of controlling animal position and keeping them away from a designated area. It worth noting that the use of sounds is different to existing approaches, which utilize electric shocks to train animals to adhere within the boundaries of a virtual fence. It was found that audio signals in the frequencies of 125Hz-440Hz, 10kHz-17kHz and white noise are able to control animal activity with accuracies of 89.88%, and 95.93%, for Hebridean and Greyface Dartmoor ewes, respectively. Last but not least, the thesis proposes a multifunctional system that identifies whether the animal is active or inactive, using transfer learning, and manipulates its position using the optimized sound settings achieving a classification accuracy of over 99.95%.
Where traditional motorways contain hard shoulders to provide refuge for broken-down vehicles, smart motorways instead use live outer lane to ease congestion. The live lanes can be closed due to ...accidents or breakdowns which is communicated to other road users through overhead gantry signs. This can only occur if the traffic management control center is made aware of the stationary vehicle(s), through either notification via phone call or Motorway Incident Detection and Automatic Signaling (MIDAS) induction loop technology. Alternatively, radar-based stopped vehicle detection is used to identify non-moving objects on the highway. However, this technology is unable to recognize objects or distinguish between congestion and break down etc., which leads to generate false alarms. For the first time, we propose a fully autonomous computer vision and deep learning-based solution to detect stationary vehicles on highways and local roads. We employ deep transfer learning to build a custom-trained vehicle detection model using a newly prepared dataset comprising over 105,000 annotated vehicle instances. DeepSort algorithm is employed for real-time vehicle tracking through associating instances between time series frames, followed by a rule-based algorithm to identify the current state of detected vehicles. Experimental outcomes show our approach as outperforming the state-of-the-art methods in terms of efficient and reliable detection of stationary vehicles (with 98.3% accuracy) as well as distinguish them from congestions when evaluated over video streams captured in realistic dynamic and diverse conditions.