[Paper Review] A systematic review of human activity recognition using smartphones.

Marcin Strączkiewicz, Jukka‐Pekka Onnela|arXiv (Cornell University)|Oct 7, 2019

Human Mobility and Location-Based Analysis56 references5 citations

TL;DR

This systematic review analyzes 72 studies on smartphone-based human activity recognition (HAR), examining data acquisition, preprocessing, feature extraction, and classification methods. It identifies prevalent practices, evaluates algorithmic variations, and outlines future research directions for improving the translation of consumer-grade smartphone data into reliable, research-grade physical activity patterns.

ABSTRACT

Smartphones have become a global communication tool and more recently a technology for studying human behavior. Given their numerous built-in sensors, smartphones are able to capture detailed and continuous observations on activities of daily living. However, translation of measurements from these consumer-grade devices into research-grade physical activity patterns remains challenging. Over the years, researchers have proposed various human activity recognition (HAR) systems which vary in algorithmic details and statistical principles. In this paper, we summarize existing approaches to smartphone-based HAR. We systematically screened the literature on Scopus, PubMed, and Web of Science in the areas of data acquisition, data preprocessing, feature extraction, and activity classification. We ultimately identified 72 articles on smartphone-based HAR. To provide an understanding of the literature, we discuss each of these areas separately, identify the most common practices and their alternatives, and propose possible future research directions for this interesting and important field.

Motivation & Objective

• To synthesize existing approaches in smartphone-based human activity recognition (HAR) across data acquisition, preprocessing, feature extraction, and classification.
• To identify the most common and effective practices in each stage of HAR using consumer-grade smartphones.
• To evaluate the strengths and limitations of current algorithms and statistical methods in HAR systems.
• To propose actionable future research directions for advancing the reliability and validity of HAR using smartphones.

Proposed method

• Systematic literature screening across Scopus, PubMed, and Web of Science focusing on smartphone-based HAR research.
• Categorization of studies based on data acquisition methods, preprocessing techniques, feature extraction strategies, and classification algorithms.
• Analysis of algorithmic diversity and statistical principles across 72 identified studies in the HAR pipeline.
• Comparison of common practices versus alternative approaches in each HAR component to assess performance and reliability.
• Synthesis of findings to highlight trends, gaps, and methodological trade-offs in smartphone-based HAR.
• Identification of research gaps and formulation of future research directions based on systematic analysis of methodological choices.

Experimental results

Research questions

• RQ1 What are the most common data acquisition methods used in smartphone-based HAR studies?
• RQ2 Which preprocessing and feature extraction techniques are most frequently applied, and how do they compare?
• RQ3 How do different classification algorithms perform across various activity recognition tasks?
• RQ4 What are the key methodological challenges in translating smartphone sensor data into research-grade physical activity patterns?
• RQ5 What future research directions are most promising for improving the validity and reliability of smartphone-based HAR?

Key findings

• The review identified 72 studies on smartphone-based HAR, highlighting significant methodological diversity across data acquisition, preprocessing, feature extraction, and classification stages.
• Accelerometer-based data collection was the most prevalent method for capturing human activity signals.
• Time-domain and frequency-domain features were the most commonly used feature extraction techniques.
• Machine learning classifiers such as SVM, Random Forest, and k-NN were frequently employed for activity classification.
• Despite widespread use, many studies lack standardized evaluation protocols, limiting comparability across research.
• The review identifies a need for more rigorous validation and standardization to improve the reliability of HAR systems for research applications.