Feature Analysis to Human Activity Recognition

Abstract

Human activity recognition (HAR) is one of those research areas whose importance and popularity have notably increased in recent years. HAR can be seen as a general machine learning problem which requires feature extraction and feature selection. In previous articles different features were extracted from time, frequency and wavelet domains for HAR but it is not clear that, how to determine the best feature combination which maximizes the performance of a machine learning algorithm. The aim of this paper is to present the most relevant feature extraction methods in HAR and to compare them with widely-used filter and wrapper feature selection algorithms. This work is an extended version of [1] a where we tested the efficiency of filter and wrapper feature selection algorithms in combination with artificial neural networks. In this paper the efficiency of selected features has been investigated on more machine learning algorithms (feed-forward artificial neural network, k-nearest neighbor and decision tree) where an independent database was the data source. The result demonstrates that machine learning in combination with feature selection can overcome other classification approaches.

Music Stimuli Recognition in Electroencephalogram Signal

“…With an appropriate feature set the classifier model will be simpler and its performance will be better [6]. In this study the features have been extracted from the time and frequency domains as in [17] where the authors collected the most relevant feature extraction methods to the human activity recognition problem. Although activity and music stimuli recognition have two different objectives, many similarities exist between them.…”

A survey on wearable sensor modality centred human activity recognition in health care

“…The number of the samples in one window versus the window size based on the reviewed works is plotted in Fig.3, with several less commonly-used numbers being excluded (Machado, et al, 2015). And we can see two obvious trends from Fig.3: one is that most sample numbers in one window fall into between 32 (Suto, et al, 2016) and 256 (Hu, et al, 2014); the other is that sample numbers of the nth power of 2 are often applied, such as 64 (Murao & Terada, 2014), and 128 (Ronao & Cho, 2016). The sampling rate as well as the trade-off between recognition efficiency and performance should be considered when manually determining the window size.…”

“…Different window sizes are employed in WSHAR, which are found to vary from 0.08s (Berchtold, et al, 2010), 0.1s (Murao & Terada, 2014), 0.2s (Zhang & Sawchuk, 2012), 0.5s , 1s (Bulling, et al, 2014), 1.6s (Suto, et al, 2016), 2s (Laudanski, et al, 2015), 2.56s (Hassan, et al, 2018), 3.88s (Chernbumroong, et al, 2014), 4s (Wang, et al, 2013, 5s (Machado, et al, 2015, 6.7s (Bao & Intille, 2004), 8.53s (Guo, et al, 2012), 9s (Kalantarian, et al, 2015), 10s (Catal, et al, 2015), 12.8s (Wang, et al, 2018) to 30s (Liu, et al, 2012) and even higher. Usually, a window covers a few seconds long time interval.…”

“…We need to take the sampling rate of sensors into account with respect to the number of samples in one window, since the sample number is determined by both the window size and the sampling rate. A wide range of sampling rates are explored in WSHAR, varying from 1hz (Zhang, et al, 2014), 5hz (Alshurafa, et al, 2014), 6hz (Gjoreski & Gams, 2011a), 10hz (Nam & Park, 2013), 20hz (Wang, et al, 2018, Suto, et al, 2016, 33hz (Chernbumroong, et al, 2014), 50hz (Biswas, et al, 2015, Hassan, et al, 2018, 64Hz (Hammerla, et al, 2016), 100hz (Sani, et al, 2017), 120hz (Laudanski, et al, 2015), 126hz (Gupta & Dallas, 2014), 135hz (Dalton & ÓLaighin, 2013), 200hz (Yao, et al, 2017), 256hz (Chen, et al, 2014), and up to 800hz (Montalto, et al, 2015). Generally, higher sampling rates can catch more signal details but coupled with higher energy requirements and higher noise impact; lower sampling rates save considerable energy, but might omit certain relevant information, thus lower accuracy.…”

Selective Feature Generation Method for Classification of Low-Dimensional Data

“…Such tools are currently being applied in various fields, including biometrics (e.g., iris, fingerprint and face recognition), data mining, diagnosis systems and pattern classification [22,26].…”