Dr Alireza Ahrabian

Research Fellow

About

Biography

Dr Alireza Ahrabian is a Research Fellow in the Department of Electronic Engineering at the University of Surrey. He completed his PhD degree in signal processing and her MEng in Electrical Engineering at Imperial College London, in 2014 and 2010, respectively.

Research interests

My main research interests are in detecting changes in time series data. Namely, my work focuses on developing both sequential and batch processing algorithms. Furthermore, I have interests in Bayesian approaches and statistical signal processing techniques.

Publications

Enshaeifar S, Hoseinitabatabaei SA, Ahrabian A, Barnaghi P (2017) Pattern Identification for State Prediction in Dynamic Data Streams,
This work proposes a pattern identification and online prediction algorithm for processing Internet of Things (IoT) time-series data. This is achieved by first proposing a new data aggregation and datadriven discretisation method that does not require data segment normalisation. We apply a dictionary based algorithm in order to identify patterns of interest along with prediction of the next pattern. The performance of the proposed method is evaluated using synthetic and real-world datasets. The evaluations results shows that our system is able to identify the patterns by up to 85% accuracy which is 16.5% higher than a baseline using the Symbolic Aggregation Approximation (SAX) method.
Ahrabian A, Kolozali S, Enshaeifar S, Cheong Took C, Barnaghi P (2017) Stream Data Analysis as a web service: A Case Study Using IoT Sensor Data,Proceedings of ICASSP2017 IEEE
DOI: 10.1109/ICASSP.2017.7953308
The advent of Internet of Things, has resulted in the development of infrastructure for capturing and storing data from domains ranging from smart devices (e.g. smartphones) to smart cities. This data is often available publicly and has enabled a wider range of data consumers to utilise such data sets for applications ranging from scientific experimentation to enhancing commercial activity for businesses. Accordingly this has resulted in the need for the development data analysis tools that are both simple to use and provide the most effective tools for a given data set. To this end, we introduce data analysis tools as web service, that enables the data consumer to make a simple HTTP request for processing data over the internet. By providing such tools as a web service, we demonstrate the potential of such a system to aid both the advanced and novice data consumer. Furthermore, this work provides an use case example of the proposed tool on publicly available data extracted from the smart city CityPulse IoT project.
Ahrabian Alireza, Elsaleh Tarek, Fathy Yasmin, Barnaghi Payam (2017) Detecting Changes in the Variance of Multi-Sensory Accelerometer Data Using MCMC,Proceedings of IEEE Sensors 2017 IEEE
DOI: 10.1109/ICSENS.2017.8234260
An important field in exploratory sensory data analysis is the segmentation of time-series data to identify activities of interest. In this work, we analyse the performance of univariate and multi-sensor Bayesian change detection algorithms in segmenting accelerometer data. In particular, we provide theoretical analysis and also performance evaluation on synthetic data and real-world data. The results illustrate the advantages of using multi-sensory variance change detection in the segmentation of dynamic data (e.g. accelerometer data).
Enshaeifar S, Farajidavar N, Ahrabian A, Barnaghi P, Hannam K, Deere K, Tobias J, Allison S (2017) Recognising Bone Loading Exercises In Older Adults Using Machine Learning,Medicine & Science in Sports & Exercise49(5S) American College of Sports Medicine
DOI: 10.1249/01.mss.0000518712.49164.ed

Machine learning has been used to accurately recognise physical activity patterns; however, classifiers for recognising targeted bone loading exercises have not been developed.

PURPOSE:

The purpose of this study was to determine the accuracy of machine learning models for classifying the intensity of exercises necessary for bone adaption in older adults.

METHODS:

Triaxial accelerometer data was collected from forty-four older participants (60-70 yrs) wearing a GCDC X16-1C accelerometer on their hip during three aerobics classes consisting of impact aerobic exercises performed at high and low intensities. Multi-class support vector machine (M-SVM) classifiers were trained in parallel for activity type detections where one classifier trained with low intensity activity samples and the other with high intensity samples. In a multi-view scoring manner, the classification confidence of these two learners was utilised for predicting the activity intensity. The leave-one-out cross-validation technique was used for assessment purpose.

RESULTS:

Overall recognition accuracy of the M-SVM classifier for detecting exercise intensity was 73%. For each aerobics class, the M-SVM classifier accurately recognised exercise intensity by 82%, 73% and 65%.

CONCLUSIONS:

Machine learning techniques such as M-SVM accurately recognised the intensity of bone promoting exercises from triaxial accelerometer data in community-dwelling older adults. First results of the developed classifier demonstrate significant potential of machine learning models for the evaluation of exercise adherence and performance in older adults.

Ahrabian Alireza, Enshaeifar Shirin, Cheong Took Clive, Barnaghi Payam (2018) Segment parameter labelling in MCMC mean-shift change detection,ICASSP 2018 IEEE
DOI: 10.1109/ICASSP.2018.8462438
This work addresses the problem of segmentation in time series data with respect to a statistical parameter of interest in Bayesian models. It is common to assume that the parameters are distinct within each segment. As such, many Bayesian change point detection models do not exploit the segment parameter patterns, which can improve performance. This work proposes a Bayesian mean-shift change point detection algorithm that makes use of repetition in segment parameters, by introducing segment class labels that utilise a Dirichlet process prior. The performance of the proposed approach was assessed on both synthetic and real world data, highlighting the enhanced performance when using parameter labelling.