Human activity data discovery from triaxial accelerometer sensor: Non-supervised learning sensitivity to feature extraction parametrization

Machado, Inês; Gomes, Ana Sofía; Gamboa, Hugo; Paixão, Vítor; Costa, Rui M.

doi:10.1016/j.ipm.2014.07.008

Cited by 67 publications

(55 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the work of Inês P. Machado et al [17], the activities of standing, sitting, walking, running, lying are studied with mixed-domain features. Four clustering methods, including K-Means, Spectral Clustering, Mean Shift and Affinity Propagation (AP) based on Euclidean distance are used to distinguish different activities.…”

Section: Related Workmentioning

confidence: 99%

Using Jaccard Distance Measure for Unsupervised Activity Recognition with Smartphone Accelerometers

Wang

et al. 2017

Web and Big Data

View full text Add to dashboard Cite

Abstract. The rapid popularity of smartphones has led to a growing research interest in human activity recognition (HAR) with the mobile devices. Accelerometer is the most commonly used sensor of smartphone for HAR. Most supervised HAR methods have been developed. However, it is very difficult to collect the annotated or labeled training data for HAR. So, developing of effective unsupervised methods for HAR is very necessary. The accuracy of the unsupervised activity recognition can be greatly affected by feature extraction methods and distance measures.Although Euclidean distance measure is commonly used in activity recognition, it is not suitable for measuring distance when the number of features is very large, which is usually the case in HAR.Jaccard distance is a distance measure based on mutual information theory and can better represent the differences between nonnegative feature vectors than Euclidean distance. In this work, the Jaccard distance measure is applied to HAR for the first time. In the experiments, the results of the Jaccard distance measure and the Eucildean distance measure are compared, using three different feature extraction methods. Two different evaluation methods are used to comprehensively analyze the final results: (a) C-Index before clustering, (b) FM-index after using five different clustering methods which are Spectral Cluster, Single-Linkage, Ward-Linkage, Average-Linkage, and K-Medoids. Experiments show that, almost for every combination of the feature extraction methods and the evaluation methods, the Jaccard distance is consistently better than the Euclidean distance for unsupervised HAR.

show abstract

Section: Related Workmentioning

confidence: 99%

Using Jaccard Distance Measure for Unsupervised Activity Recognition with Smartphone Accelerometers

Wang

et al. 2017

Web and Big Data

View full text Add to dashboard Cite

show abstract

“…Activity recognition is a way of using sensors to identify activities and provide sensor context awareness to diverse applications [9]. To achieve high accuracy in activity recognition tasks, researchers like Machado et al [10] and Roggen et al [11] used expensive methodologies such as human observers and video labelling respectively. However, these solutions cannot be adapted for automated labelling because of their dependence on third party labellers.…”

Section: A Sensor Data Labellingmentioning

confidence: 99%

A Gamification Framework for Sensor Data Analytics

L’Heureux

Grolinger

Higashino

et al. 2017

2017 IEEE International Congress on Internet of Things (ICIOT)

View full text Add to dashboard Cite

Abstract-The Internet of Things (IoT) enables connected objects to capture, communicate, and collect information over the network through a multitude of sensors, setting the foundation for applications such as smart grids, smart cars, and smart cities. In this context, large scale analytics is needed to extract knowledge and value from the data produced by these sensors. The ability to perform analytics on these data, however, is highly limited by the difficulties of collecting labels. Indeed, the machine learning techniques used to perform analytics rely upon data labels to learn and to validate results. Historically, crowdsourcing platforms have been used to gather labels, yet they cannot be directly used in the IoT because of poor human readability of sensor data. To overcome these limitations, this paper proposes a framework for sensor data analytics which leverages the power of crowdsourcing through gamification to acquire sensor data labels. The framework uses gamification as a socially engaging vehicle and as a way to motivate users to participate in various labelling tasks. To demonstrate the framework proposed, a case study is also presented. Evaluation results show the framework can successfully translate gamification events into sensor data labels.

show abstract

“…It is highly important to make comparisons of chronological series which appear, very often, deformed in time by presenting a similarity of amplitude and forms with a phase shift in the time. Time shift problem is often encountered in the analysis of data obtained from biological experiments Aach and Church (2001); Clifford et al (2009), online signature validation Xia et al (2018), pattern recognition Keogh and Ratanamahatana (2005); Ratanamahatana and Keogh (2004); Ratanamahatana et al (2010); Araújo et al (2015) as well as in signal processing Barth et al (2015) and the recognition of human activities Machado et al (2015); Folgado et al (2018). The clear majority of approaches applied in this area have been focusing on indexing based on the Euclidean metric Agrawal et al (1995); Chan et al (2003); Das et al (1998); Debregeas and Hebrail (1998) ;Faloutsos et al (1994); Keogh et al (2001); Korn et al (1997); Yi and Faloutsos (2000), which assumes that discrete signals are equidistant points in time and are also aligned in time axis.…”

Section: Introductionmentioning

confidence: 99%

An Automated Time-Shift Alignment Algorithm Based on Discrete to Continuous Approach

Rhalem¹,

Raji²,

Hammouch³

et al. 2019

Journal of Computer Science

View full text Add to dashboard Cite

Dealing with the variation in retention time and time series is one of the most popular challenges in the field of analysing data obtained from biological experiments. In most cases, this problem can lead to invalid conclusions. This work aims to propose a new method to cope with this problem by using the principle of the transition from discrete to continuous entities. The experimental results have shown good performance of the proposed algorithm that has been applied on different types of time series data.

show abstract

Human activity data discovery from triaxial accelerometer sensor: Non-supervised learning sensitivity to feature extraction parametrization

Cited by 67 publications

References 15 publications

Using Jaccard Distance Measure for Unsupervised Activity Recognition with Smartphone Accelerometers

Using Jaccard Distance Measure for Unsupervised Activity Recognition with Smartphone Accelerometers

A Gamification Framework for Sensor Data Analytics

An Automated Time-Shift Alignment Algorithm Based on Discrete to Continuous Approach

Contact Info

Product

Resources

About