Latent Independent Excitation for Generalizable Sensor-based Cross-Person Activity Recognition

Qian, Hangwei; Pan, Sinno Jialin; Chen, Miao

doi:10.1609/aaai.v35i13.17416

Cited by 42 publications

(13 citation statements)

References 56 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Therefore, the deep learning architectures for interpreting ECG and PPG signals may have to be optimized for learning from quasi-periodic signals, as well as for regression. Moreover, applying restrictions on the feature maps or embeddings extracted by the deep learning models, such as the usage of domain adversarial training ( 28 ) or generalizable independent latent excitation ( 51 ) can help enforcing subject non-specificity of the model. These methods may improve the cross-subject generalization capability of BP estimation models.…”

Section: Resultsmentioning

confidence: 99%

PulseDB: A large, cleaned dataset based on MIMIC-III and VitalDB for benchmarking cuff-less blood pressure estimation methods

Wang

Mohseni²,

Kilgore³

et al. 2023

Front. Digit. Health

View full text Add to dashboard Cite

There has been a growing interest in developing cuff-less blood pressure (BP) estimation methods to enable continuous BP monitoring from electrocardiogram (ECG) and/or photoplethysmogram (PPG) signals. The majority of these methods have been evaluated using publicly-available datasets, however, there exist significant discrepancies across studies with respect to the size, the number of subjects, and the applied pre-processing steps for the data that is eventually used for training and testing the models. Such differences make conducting performance comparison across models largely unfair, and mask the generalization capability of various BP estimation methods. To fill this important gap, this paper presents “PulseDB,” the largest cleaned dataset to date, for benchmarking BP estimation models that also fulfills the requirements of standardized testing protocols. PulseDB contains 1) 5,245,454 high-quality 10-s segments of ECG, PPG, and arterial BP (ABP) waveforms from 5,361 subjects retrieved from the MIMIC-III waveform database matched subset and the VitalDB database; 2) subjects’ identification and demographic information, that can be utilized as additional input features to improve the performance of BP estimation models, or to evaluate the generalizability of the models to data from unseen subjects; and 3) positions of the characteristic points of the ECG/PPG signals, making PulseDB directly usable for training deep learning models with minimal data pre-processing. Additionally, using this dataset, we conduct the first study to provide insights about the performance gap between calibration-based and calibration-free testing approaches for evaluating generalizability of the BP estimation models. We expect PulseDB, as a user-friendly, large, comprehensive and multi-functional dataset, to be used as a reliable source for the evaluation of cuff-less BP estimation methods.

show abstract

Section: Resultsmentioning

confidence: 99%

PulseDB: A large, cleaned dataset based on MIMIC-III and VitalDB for benchmarking cuff-less blood pressure estimation methods

Wang

Mohseni²,

Kilgore³

et al. 2023

Front. Digit. Health

View full text Add to dashboard Cite

show abstract

“…For fall detection (Fall-8), the model achieved an accuracy of 55.20%, an AUC of 67.82%, and an F1-score of 55%. A variational AE is proposed for HAR using the UniMiB-SHAR (Full-17) dataset in [10]. The latter achieves an average performance of 24.01% on the Full-17 dataset.…”

Section: Related Workmentioning

confidence: 99%

Human Activity Recognition Using Convolutional Autoencoder and Advanced Preprocessing

Zaoui,

Benabbou,

Ettaoufik

et al. 2024

Int. J. Onl. Eng.

View full text Add to dashboard Cite

E-health systems rely on information and communication technology to support and improve various aspects of health services, delivery, and management. The success of artificial intelligence techniques has led to the emergence of a variety of systems designed to address a wide range of healthcare issues. In particular, gathering data on patient activity and behavior has enabled the development of reliable predictive systems for detecting chronic diseases and forecasting their progression. Human activity detection is a vast and emerging field, and various datasets have been collected for training different machine learning and deep learning (DL) models. The University of Milano Bicocca smartphone-based human activity recognition (UniMiB-SHAR) dataset is widely used for analyzing and recognizing human actions, including walking, running, and other daily activities. However, the autoencoder (AE) technique trained on this dataset yields poor performance. This paper aims to enhance the performance of AEs on the challenging UniMiB-SHAR dataset by introducing a convolutional AE model and employing novel preprocessing techniques, including normalization, magnitude, principal component analysis (PCA), and balancing methods such as SMOTEEN and ADASYNE. The experimental results demonstrate that the proposed AE model achieved successful performance, surpassing the state-of-the-art methods, with accuracies of 96.56% for activities of daily living (ADL), 98.86% for Fall, and 88.47% for the full dataset.

show abstract

“…Thus, SDA blurs the distribution information of different users when decreasing the intra-class distance. However, even for the same target user, the contribution of utilizing different users as source domains is also different, implying that the distribution similarity is disparate between different subjects and has to be considered separately [24][25][26][27][28]. SDA blurs all source domains as one source which is a limitation to the performance of SDA.…”

Section: Introductionmentioning

confidence: 99%

Multi-source domain generalization and adaptation toward cross-subject myoelectric pattern recognition

Zhang

et al. 2023

J. Neural Eng.

View full text Add to dashboard Cite

Myoelectric pattern recognition (MPR) has shown satisfactory performance under ideal laboratory conditions. Nevertheless, the individual variances lead to dramatic performance degradation in cross-user MPR applications. It is crucial to enable the myoelectric interface to adapt to multiple users’ surface electromyography (sEMG) distributions in practical. Domain adaptation (DA) is a promising approach to tackle cross-user challenges due to its ability to diminish the divergence between individual users' EMG distributions and escalate model generalization performance. However, existing DA methods in sEMG control are based on single-source domain adaptation (SDA). SDA solely mixes multiple training users’ data as a combined source domain and attempts to align with a novel user. This simple data mixing manner ignores the sEMG distribution variations between disparate training users, leading to an insufficient variance elimination and lower performance. To this end, this paper proposes a multi-source synchronize domain adaptation (MS-DA) framework with both domain adaptation and domain generalization (DG) capability. This multi-source framework aligns each source user and the new user in individual feature spaces, which better transfers the knowledge of existing users to the new user. Moreover, we retain the source-combined data to preserve the effectiveness of SDA. The property was further confirmed by evaluating the performance of the proposed method on data from nine subjects performing six tasks. Experiment results prove that the proposed multi-source framework achieved both positive DG and DA performance in a cross-user classification manner. This work demonstrates the usability and feasibility of the proposed multi-source framework in cross-user myoelectric control.

show abstract

Latent Independent Excitation for Generalizable Sensor-based Cross-Person Activity Recognition

Cited by 42 publications

References 56 publications

PulseDB: A large, cleaned dataset based on MIMIC-III and VitalDB for benchmarking cuff-less blood pressure estimation methods

PulseDB: A large, cleaned dataset based on MIMIC-III and VitalDB for benchmarking cuff-less blood pressure estimation methods

Human Activity Recognition Using Convolutional Autoencoder and Advanced Preprocessing

Multi-source domain generalization and adaptation toward cross-subject myoelectric pattern recognition

Contact Info

Product

Resources

About