Deep-Learning-Based Signal Enhancement of Low-Resolution Accelerometer for Fall Detection Systems

Abstract

In the last two decades, fall detection (FD) systems have been developed as a popular assistive technology. To support long-term FD services, various power-saving strategies have been implemented. Among them, a reduced sampling rate is a common approach for an energy-efficient system in the real world. However, the performance of FD systems is diminished owing to low-resolution (LR) accelerometer signals. To improve the detection accuracy with LR accelerometer signals, several technical challenges must be considered, including mismatch of effective features and the degradation effects. In this work, a deeplearning-based accelerometer signal enhancement (ASE) model is proposed as a front-end processor to help typical LR-FD systems achieve better detection performance. The proposed ASE model based on a deep denoising convolutional autoencoder architecture reconstructs high-resolution (HR) signals from the LR signals by learning the relationship between the LR and HR signals. The results show that the FD system using support vector machine and the proposed ASE model at an extremely low sampling rate (sampling rate < 2 Hz) achieved 97.34% and 90.52% accuracies in the SisFall and FallAllD datasets, respectively, while those without ASE models only achieved 95.92% and 87.47% accuracies in the SisFall and FallAllD datasets, respectively. The results also demonstrate that the proposed ASE mode can be suitably combined with deep-learning-based FD systems.

SRECG: ECG Signal Super-Resolution Framework for Portable/Wearable Devices in Cardiac Arrhythmias Classification

“…The experimental results validated the DL approach for low-resolution CAs classification [24]. However, these previous studies did not investigate automatic CAs classification at very low sampling rates, while several studies have developed P/W-based monitoring systems with an extremely low sampling frequency to support other long-term healthcare services, such as daily activity recognition and fall detection [10,25].…”

“…However, there are certain limits tied to the current capability of the hardware, such as battery life and computing power, for ECG long-term monitoring systems implemented in CE. With respect to the limitation on energy consumption, it is understood that more power is required when a higher frequency is used to record ECG signals [9][10][11]. Therefore, in this study, we attempted to alleviate this problem by recovering the predictive power of the HMC with a low sampling frequency input via the proposed DL-based SRECG framework, which is a novel SR-based method.…”

“…frequencies, up to 90% of power consumption of health monitoring systems is due to data sampling [9][10][11].…”

“…Various power efficiency strategies have been applied to automatic CAs classification systems to enable long-term health monitoring services [10][11][12]. Maji et al proposed an adaptive power management approach to classify critical and non-critical signals before further DL-based ECG classification [13].…”

Research on concentration detection algorithm of embedded gas detection system based on deep learning

“…These toxic gases pose a serious threat to human life and health [3]. In recent years, the incidence of lung cancer has increased year by year, which is not unrelated to these poisonous gases [4]. In production activities, due to irregular production processes and cost pressures, such as oil production process problems, the formation of haze weather is directly caused [5].…”