A cross-dataset deep learning-based classifier for people fall detection and identification

Delgado-Escaño, Rubén; Castro, Francisco M.; Cózar, Julián Ramos; Marín-Jiménez, Manuel J.; Guil, Nicolás; Casilari, E.

doi:10.1016/j.cmpb.2019.105265

Cited by 42 publications

(32 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…On the other hand, Ivascu et al[55] and Casilari et al[56] both used an ANN-based model achieving an accuracy of 96.73% and 91.09%, respectively. Delgado-Escaño et al[57] utilized KNN ML algorithm where the results exceeded the ones mentioned earlier. Ours is simulated over the UniMiB-SHAR dataset where the results portray ENN as the best classifier over the 90/10 train/test ratio averaged over the 10-fold cross-validation, taken from Table7.…”

mentioning

confidence: 87%

An IoT-based framework for remote fall monitoring

Al‐Kababji

Amira

Bensaali

et al. 2021

Biomedical Signal Processing and Control

View full text Add to dashboard Cite

Fall detection is a serious healthcare issue that needs to be solved. Falling without quick medical intervention would lower elderly's chances of survival, especially if living alone. Hence, the need is there for developing fall detection algorithms with high accuracy. This paper presents a novel IoT-based system for fall detection that includes a sensing device transmitting data to a mobile application through a cloud-connected gateway device. Then, the focus is shifted to the algorithmic aspect where multiple features are extracted from 3-axis accelerometer data taken from existing datasets. The results emphasize on the significance of Continuous Wavelet Transform (CWT) as an influential feature for determining falls. CWT, Signal Energy (SE), Signal Magnitude Area (SMA), and Signal Vector Magnitude (SVM) features have shown promising classification results using K-Nearest Neighbors (KNN) and E-Nearest Neighbors (ENN). For all performance metrics (accuracy, recall, precision, specificity, and F 1 Score), the achieved results are higher than 95% for a dataset of small size, while more than 98.47% score is achieved in the aforementioned criteria over the UniMiB-SHAR dataset by the same algorithms, where the classification time for a single test record is extremely efficient and is real-time.

show abstract

mentioning

confidence: 87%

An IoT-based framework for remote fall monitoring

Al‐Kababji

Amira

Bensaali

et al. 2021

Biomedical Signal Processing and Control

View full text Add to dashboard Cite

show abstract

“…However, previous studies analyzed completely different datasets for an effective test, representing here as a cross-dataset. Cao et al [ 21 ] and Delgado-Escaño et al [ 22 ] used different datasets for training and testing to evaluate the algorithm. Cao et al [ 21 ] suggested the adaptive action detection algorithm from human video with high accuracy (95.02%) and used four different datasets to generalize the action detection model.…”

Section: Introductionsmentioning

confidence: 99%

“…Cao et al [ 21 ] suggested the adaptive action detection algorithm from human video with high accuracy (95.02%) and used four different datasets to generalize the action detection model. Delgado-Escaño et al [ 22 ] presented a new cross-dataset classifier based on a deep architecture and a k-NN classifier for fall detection and people identification. They tested their algorithm using four different public IMU datasets.…”

Section: Introductionsmentioning

confidence: 99%

The Performance of Post-Fall Detection Using the Cross-Dataset: Feature Vectors, Classifiers and Processing Conditions

Koo

Kim

Nam

et al. 2021

Sensors

View full text Add to dashboard Cite

In this study, algorithms to detect post-falls were evaluated using the cross-dataset according to feature vectors (time-series and discrete data), classifiers (ANN and SVM), and four different processing conditions (normalization, equalization, increase in the number of training data, and additional training with external data). Three-axis acceleration and angular velocity data were obtained from 30 healthy male subjects by attaching an IMU to the middle of the left and right anterior superior iliac spines (ASIS). Internal and external tests were performed using our lab dataset and SisFall public dataset, respectively. The results showed that ANN and SVM were suitable for the time-series and discrete data, respectively. The classification performance generally decreased, and thus, specific feature vectors from the raw data were necessary when untrained motions were tested using a public dataset. Normalization made SVM and ANN more and less effective, respectively. Equalization increased the sensitivity, even though it did not improve the overall performance. The increase in the number of training data also improved the classification performance. Machine learning was vulnerable to untrained motions, and data of various movements were needed for the training.

show abstract

“…However, the impact of cross-position on detection performance has not been fully explored and only one public dataset is validated in their works. Delgado-Escan õ et al [12] developed a cross-dataset deep learning-based fall detector for tackling cross-configuration problems. They employed multi-task learning approaches and CNN-LSTM models to obtain effective features that can perform well in various datasets.…”

Section: Introductionmentioning

confidence: 99%

“…Moreover, it is difficult to collect the labeled data from all possible sensor positions. (ii) Cross-configuration: The distribution of the recorded patterns is heterogeneous under different hardware conditions (e.g., sampling rate, sensing range, resolution, and noise density), as various commercial sensors are available in the wearable device market [12]. For example, the model trained on the smartphone is unsuitable for a direct deployment on smartwatches because of the possible differences in sampling rate, sensing range, and resolutions.…”

Section: Introductionmentioning

confidence: 99%

Domain-Adaptive Fall Detection Using Deep Adversarial Training

Liu

Chan

Kuo

et al. 2021

IEEE Trans. Neural Syst. Rehabil. Eng.

View full text Add to dashboard Cite

Fall detection (FD) systems are important assistive technologies for healthcare that can detect emergency fall events and alert caregivers. However, it is not easy to obtain large-scale annotated fall events with various specifications of sensors or sensor positions during the implementation of accurate FD systems. Moreover, the knowledge obtained through machine learning has been restricted to tasks in the same domain. The mismatch between different domains might hinder the performance of FD systems. Cross-domain knowledge transfer is very beneficial for machine-learning based FD systems to train a reliable FD model with well-labeled data in new environments. In this study, we propose domain-adaptive fall detection (DAFD) using deep adversarial training (DAT) to tackle cross-domain problems, such as cross-position and cross-configuration. The proposed DAFD can transfer knowledge from the source domain to the target domain by minimizing the domain discrepancy to avoid mismatch problems. The experimental results show that the average F1-score improvement when using DAFD ranges from 1.5% to 7% in the cross-position scenario, and from 3.5% to 12% in the cross-configuration scenario, compared to using the conventional FD model without domain adaptation training. The results demonstrate that the proposed DAFD successfully helps to deal with cross-domain problems and to achieve better detection performance.

show abstract

A cross-dataset deep learning-based classifier for people fall detection and identification

Cited by 42 publications

References 16 publications

An IoT-based framework for remote fall monitoring

An IoT-based framework for remote fall monitoring

The Performance of Post-Fall Detection Using the Cross-Dataset: Feature Vectors, Classifiers and Processing Conditions

Domain-Adaptive Fall Detection Using Deep Adversarial Training

Contact Info

Product

Resources

About