Conference Papers

Manta: Enhancing Mamba for Few-Shot Action Recognition of Long Sub-Sequence

An intsereting work on few-shot action recognition with Mamba, which is a challenging task in videos of long sequences.

Wenbo Huang, Jinghui Zhang*, Guang Li, Lei Zhang, Shoyuan Wang, Fang Dong, Jiahui Jin, Takahiro Ogawa, Miki Haseyama

AAAI 2025-1 (Conference Paper)

2025 the 39th Annual AAAI Conference on Artificial Intelligence, Philadelphia, USA (CCF Rank A, Unknown Presentation, Accept Rate 23.4%)

PDF PDF (Full edition) Poster Code (GitHub) OpenReview Cite by bib

SOAP: Enhancing Spatio-Temporal Relation and Motion Information Capturing for Few-Shot Action Recognition

An intsereting work on few-shot action recognition, which is a challenging task in high frame-rate videos.

Wenbo Huang, Jinghui Zhang*, Xuwei Qian, Zhen Wu, Meng Wang, Lei Zhang

MM 2024-1 (Conference Paper)

2024 the 32nd ACM International Conference on Multimedia, Melbourne, Australia (CCF Rank A, Poster Presentation, Accept Rate 26.2%)

PDF PDF (Full edition) Poster Code (GitHub) OpenReview Cite by bib

Generalizable Sensor-Based Activity Recognition via Categorical Concept Invariant Learning

An intsereting work about sensor-based human activity recognition, focusing on categorical concept invariant learning.

Di Xiong, Shoyuan Wang, Lei Zhang*, Wenbo Huang, Chaolei Han

AAAI 2025-2 (Conference Paper)

2025 the 39th Annual AAAI Conference on Artificial Intelligence, Philadelphia, USA (CCF Rank A, Unknown Presentation, Accept Rate 23.4%)

PDF PDF (Full edition) Poster Code (GitHub) OpenReview Cite by bib

E2E-MFD: Towards End-to-End Synchronous Multimodal Fusion Detection

An amazimg work on multimodal fusion detection.

Jiaqing Zhang, Mingxiang Cao, Weiying Xie*, Jie Lei, Daixun Li, Geng Yang, Wenbo Huang, Yunsong Li, Xue Yang*

NeurIPS 2024-1 (Conference Paper)

2024 the 38th Annual Conference on Neural Information Processing Systems, Vancouver, Canada (CCF Rank A, Oral Presentation, Accept Rate 25.8%)

PDF Code (GitHub) Cite by bib

Journal Papers

Channel-Equalization-HAR: A Light-weight Convolutional Neural Network for Wearable Sensor Based Human Activity Recognition

We provide an alternative called Channel Equalization in this paper to reactivate these inhibited channels by performing whitening or decorrelation operation, which compels all channels to contribute more or less to feature representation.

Wenbo Huang , Lei Zhang*, Hao Wu, Fuhong Min, Aiguo Song

TMC 2022-1 (Journal Paper)

IEEE Transactions on Mobile Computing (CCF Rank A, IF=7.9, ESI Top 1% Highly Cited)

DOI: 10.1109/TMC.2022.3174816

PDF Code (GitHub) Code (In Jittor) Website Online 中译版 Cite by bib

Deep Ensemble Learning for Human Activity Recognition Using Wearable Sensors via Filter Activation

We propose a novel CNN for HAR that uses filter activation. In comparison with filter pruning that is motivated for efficient consideration, filter activation aims to activate these invalid filters from an accuracy boosting perspective.

Wenbo Huang , Lei Zhang*, Shuoyuan Wang, Hao Wu, Aiguo Song

TECS 2022-1 (Journal Paper)

ACM Transactions on Embedded Computing Systems (CCF Rank B, IF=2.0, ESI Top 1% Highly Cited)

DOI: 10.1145/3551486

PDF Code (GitHub) Code (In Jittor) Website Online 中译版 Cite by bib

The Convolutional Neural Networks Training with Channel-Selectivity for Human Activity Recognition Based on Sensors

We use the Channel-selectivity method to greatly improve the classification accuracy of the sensor signals by CNN without increasing the parameters.

Wenbo Huang , Lei Zhang*, Qi Teng, Chaoda Song, Jun He

JBHI 2021-1 (Journal Paper)

IEEE Journal of Biomedical and Health Informatics (Old Name：IEEE Transactions on Information Technology in Biomedicine, CCF Rank C, IF=7.7)

DOI: 10.1109/JBHI.2021.3092396

PDF Code (GitHub) Code (In Jittor) Website Online 中译版 Cite by bib

Shallow Convolutional Neural Networks for Human Activity Recognition using Wearable Sensors

We pay more attention to Cross-channel-communication and apply it on CNN. This method can make shallow CNNs have a similar performance with deep CNNs when classify sensor singal.

Wenbo Huang, Lei Zhang*, Wenbin Gao , Fuhong Min, Jun He

TIM 2021-1 (Journal Paper)

IEEE Transactions on Instrumentation and Measurement (CIS Rank T1, CAA Rank B, IF=5.6)

DOI: 10.1109/TIM.2021.3091990

PDF Code (GitHub) Code (In Jittor) Website Online 中译版 Cite by bib

Deep Neural Networks for Sensor-Based Human Activity Recognition Using Selective Kernel Convolution

We use Selective Kernal Convolution on CNN. This method improves the performance of deep neural networks.

Wenbin Gao, Lei Zhang*, Wenbo Huang, Fuhong Min, Jun He, Aiguo Song

TIM 2021-2 (Journal Paper)

IEEE Transactions on Instrumentation and Measurement (CIS Rank T1, CAA Rank B, IF=5.6)

DOI: 10.1109/TIM.2021.3102735

PDF Code (GitHub) Website Online Cite by bib

Deep Convolutional Networks with Tunable Speed-Accuracy Trade-off for Human Activity Recognition Using Wearables

We use Tunable Speed-Accuracy Trade-off on CNN. This method can complete Human Activity Recognition tasks more effectiviely.

Xing Wang, Lei Zhang*, Wenbo Huang, Shuoyuan Wang, Jun He, Aiguo Song

TIM 2021-3 (Journal Paper)

IEEE Transactions on Instrumentation and Measurement (CIS Rank T1, CAA Rank B, IF=5.6)

DOI: 10.1109/TIM.2021.3132088

PDF Code (GitHub) Website Online Cite by bib

Deformable Convolutional Networks for Multimodal Human Activity Recognition using Wearable Sensors

Deformable Convolutional Networks for Multimodal Human Activity Recognition.

Suige Xu, Lei Zhang*, Wenbo Huang, Hao Wu, Aiguo Song

TIM 2022-1 (Journal Paper)

IEEE Transactions on Instrumentation and Measurement (CIS Rank T1, CAA Rank B, IF=5.6)

DOI: 10.1109/TIM.2022.3158427

PDF Code (GitHub) Website Online Cite by bib

PatchHAR: A MLP-like Architecture for Efficient Activity Recognition using Wearables

We propose a lightweight network architecture called all-MLP for HAR, which is entirely built on MLP layers with a gating unit.

Shuoyuan Wang, Lei Zhang*, Xing Wang, Wenbo Huang, Hao Wu, Aiguo Song

TBIOM-2024-1 (Journal Paper)

IEEE Transactions on Biometrics, Behavior, and Identity Science

10.1109/TBIOM.2024.3354261

PDF Code (GitHub) Website Online Cite by bib

Human Activity Recognition Using Wearable Sensors by Heterogeneous Convolutional Neural Networks

We use Heterogeneous Convolutional Neural Networkse on Human Activity Recognition tasks.

Chaolei Han, Lei Zhang*, Yin Tang, Wenbo Huang, Fuhong Min, Jun He

ESWA 2022-1 (Journal Paper)

Elesvier Expert Systems with Applications (CCF Rank C, IF=8.5)

DOI: 10.1016/J.ESWA.2022.116764

PDF Code (GitHub) Website Online Cite by bib

 

Theses

Research on Multiple-Channel Convolutional Neural Network based Human Activity Recognition

Learning the influence of multiple channels on CNN based HAR tasks. Because of some people’s obstruction, this work process is very bumpy.

Wenbo Huang

Master Thesis, School of Electrical and Automation Engineering, Nanjing Normal University; June, 2022

PDF Slide