Human Activity Recognition: A Comparative Study to Assess the Contribution Level of Accelerometer, ECG, and PPG Signals

Abstract

Inertial sensors are widely used in the field of human activity recognition (HAR), since this source of information is the most informative time series among non-visual datasets. HAR researchers are actively exploring other approaches and different sources of signals to improve the performance of HAR systems. In this study, we investigate the impact of combining bio-signals with a dataset acquired from inertial sensors on recognizing human daily activities. To achieve this aim, we used the PPG-DaLiA dataset consisting of 3D-accelerometer (3D-ACC), electrocardiogram (ECG), photoplethysmogram (PPG) signals acquired from 15 individuals while performing daily activities. We extracted hand-crafted time and frequency domain features, then, we applied a correlation-based feature selection approach to reduce the feature-set dimensionality. After introducing early fusion scenarios, we trained and tested random forest models with subject-dependent and subject-independent setups. Our results indicate that combining features extracted from the 3D-ACC signal with the ECG signal improves the classifier’s performance F1-scores by 2.72% and 3.00% (from 94.07% to 96.80%, and 83.16% to 86.17%) for subject-dependent and subject-independent approaches, respectively.

Exploring the Possibility of Photoplethysmography-Based Human Activity Recognition Using Convolutional Neural Networks

“…Thus, more features, such as heartbeat, can be learned. However, an increase in window size did not necessarily increase accuracy [ 44 , 45 , 46 ]. To determine the optimal input signal length, i.e., window size, we segmented the PPG signal into lengths ranging from 2 to 20 s in increments of 2 s. The accuracy increased with increasing window sizes but converged drastically as the window size reached 10 s or more.…”

“…The effect of the input signal length on the HAR system performance has been investigated to determine the “optimal” or “cut-off” window size [ 43 , 44 ]. The optimal window size significantly varies according to parameters such as signal type, number of class categories, and activity type [ 45 , 46 ].…”

BIOWISH: Biometric Recognition Using Wearable Inertial Sensors Detecting Heart Activity

“…The design of techniques able to carry out highly efficient SCG-and GCG-based HAR in scenarios with a high number of possible ADL is therefore a research topic that could bring relevant benefits for biometric recognition applications too. About this, it is worth remarking that HAR can be performed much more efficiently when using accelerometer data, with respect to alternatives comprising signals such as ECG or PPG [66], an advantage that should further support the interest in mechanical measurements of the heart activity for biometric recognition purposes.…”

Avaliação De Medidas Para Recuperação De Curvas Sobre Movimentos Corporais

“…Diversas medidas e métodos são encontrados na literatura para analisar e comparar séries temporais de movimentos do corpo humano (XU;HE;ZHANG, 2019;ARANI;COSTA;SHIHAB, 2021;LAN, 2021;YOSHIHI et al, 2021;LI;WANG, 2022). Nesses trabalhos, a principal abordagem adotada é o estudo de séries temporais obtidas por meio de sensores inerciais, como o acelerômetro, o giroscópio e o magnetômetro, os quais são acoplados ao corpo humano para obter dados procedentes da atividade realizada pelo usuário.…”

“…Nesse cenário, sensores inerciais acoplados ao corpo são amplamente utilizados com o objetivo de obter dados de atividades corporais, possibilitando aplicações como o monitoramento do processo de reabilitação, avaliação e diagnóstico de lesões e de limitações de pacientes (CORNACCHI et al, 2017;ARANI;COSTA;SHIHAB, 2021;QIU et al, 2022).…”