Gender Recognition Based on Gradual and Ensemble Learning from Multi-View Gait Energy Images and Poses
Abstract
:1. Introduction
- The GEI provides a concise representation of movement that can be used for gender classification. However, the GEI lacks photometric information and does not clearly display body shapes. We observe that postures, such as stance and swing images of the walking cycle, exhibit unique features that can provide complementary information for gender classification. In order to improve the gender classification accuracy, we exploit multiple modality inputs of the GEI and postures.
- We propose a multi-stream network for feature extraction from the multiple modality inputs. The extracted features are fused and fed to the classifier. We design the training process to allow the feature extractor network to gradually learn from a variety of inputs. The proposed cascade framework, through ensemble learning, predicts the gender class irrespective of other factors such as viewing angle and walking status.
- We adopt data augmentation to address the class imbalance problem of the gait dataset. An investigation is performed on the CASIA B and OU-ISIR MVLP datasets. Comparison analysis is carried out with recently proposed methods based on deterministic and deep learning approaches. We demonstrate that our proposed models outperform these reference methods that only utilize either the GEI or posture image.
2. Related Work
2.1. Deterministic Algorithm
2.2. Deep Learning Model
2.3. Gait Dataset
3. Gender Classification Framework
3.1. Gait Feature Extraction from GEI
3.2. Gait Features Extraction from Stance and Swing Images
3.3. Gender Prediction CNN
3.4. Training Process
4. Experiments and Results
5. Ablation Study
6. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
Appendix A
Training Set | Number of Samples |
---|---|
0° + 18° | 3716 |
0° + 36° | 3691 |
0° + 54° | 3704 |
0° + 72° | 3695 |
0° + 90° | 3684 |
0° + 108° | 3727 |
0° + 126° | 3678 |
0° + 144° | 3679 |
0° + 162° | 3700 |
0° + 180° | 3705 |
18° + 36° | 3717 |
18° + 54° | 3730 |
18° + 72° | 3721 |
18° + 90° | 3710 |
18° + 108° | 3753 |
18° + 126° | 3704 |
18° + 144° | 3705 |
18° + 162° | 3726 |
18° + 180° | 3731 |
36° + 54° | 3705 |
36° + 72° | 3696 |
36° + 90° | 3685 |
36° + 108° | 3728 |
36° + 126° | 3679 |
36° + 144° | 3680 |
36° + 162° | 3701 |
36° + 180° | 3706 |
54° + 72° | 3709 |
54° + 90° | 3698 |
54° + 108° | 3741 |
54° + 126° | 3692 |
54° + 144° | 3693 |
54° + 162° | 3714 |
54° + 180° | 3719 |
72° + 90° | 3689 |
72° + 108° | 3732 |
72° + 126° | 3683 |
72° + 144° | 3684 |
72° + 162° | 3705 |
72° + 180° | 3710 |
90° + 108° | 3721 |
90° + 126° | 3672 |
90° + 144° | 3673 |
90° + 162° | 3694 |
90° + 180° | 3699 |
108° + 126° | 3715 |
108° + 144° | 3716 |
108° + 162° | 3737 |
108° + 180° | 3742 |
126° + 144° | 3667 |
126° + 162° | 3688 |
126° + 180° | 3693 |
144° + 162° | 3689 |
144° + 180° | 3694 |
162° + 180° | 3715 |
References
- Foggia, P.; Greco, A.; Saggese, A.; Vento, M. Multi-task learning on the edge for effective gender, age, ethnicity and emotion recognition. Eng. Appl. Artif. Intell. 2023, 118, 105651. [Google Scholar] [CrossRef]
- Azhar, M.; Ullah, S.; Raees, M.; Rahman, K.U.; Rehman, I.U. A real-time multi-view gait-based automatic gender classification system using Kinect sensor. Multimed. Tools Appl. 2022, 82, 11993–12016. [Google Scholar] [CrossRef]
- Guffanti, D.; Brunete, A.; Hernando, M. Non-invasive multi-camera gait analysis system and its application to gender classification. IEEE Access 2020, 8, 95734–95746. [Google Scholar] [CrossRef]
- Greco, A.; Saggese, A.; Vento, M. Digital signage by real-time gender recognition from face images. In Proceedings of the IEEE International Workshop on Metrology for Industry 4.0 & IoT, Roma, Italy, 3–5 June 2020; pp. 309–313. [Google Scholar]
- Duan, M.; Li, K.; Yang, C.; Li, K. A hybrid deep learning CNN-ELM for age and gender classification. Neurocomputing 2018, 275, 448–461. [Google Scholar] [CrossRef]
- Fang, J.; Yuan, Y.; Lu, X.; Feng, Y. Multi-stage learning for gender and age prediction. Neurocomputing 2019, 334, 114–124. [Google Scholar] [CrossRef]
- Amsaprabhaa, M.; Nancy, J.Y.; Khanna, N.H. A survey on spatio-temporal framework for kinematic gait analysis in RGB videos. J. Vis. Commun. Image Represent. 2021, 79, 103218. [Google Scholar] [CrossRef]
- Harris, E.J.; Khoo, I.-H.; Demircan, E. A survey of human gait-based artificial intelligence applications. Front. Robot. AI 2022, 8, 749274. [Google Scholar] [CrossRef]
- Liao, R.; Yu, S.; An, W.; Huang, Y. A model-based gait recognition method with body pose and human prior knowledge. Pattern Recognit. 2020, 98, 107069. [Google Scholar] [CrossRef]
- Han, J.; Bhanu, B. Individual recognition using gait energy image. IEEE Trans. Pattern Anal. Mach. Intell. 2006, 28, 316–322. [Google Scholar] [CrossRef]
- Luo, J.; Tjahjadi, T. Gait recognition and understanding based on hierarchical temporal memory using 3D gait semantic folding. Sensors 2020, 20, 1646. [Google Scholar] [CrossRef]
- Kovač, J.; Štruc, V.; Peer, P. Frame-based classification for cross-speed gait recognition. Multimed. Tools Appl. 2019, 78, 5621–5643. [Google Scholar] [CrossRef]
- El-Alfy, E.-S.M.; Binsaadoon, A.G. Automated gait-based gender identification using fuzzy local binary patterns with tuned parameters. J. Ambient. Intell. Humaniz. Comput. 2019, 10, 2495–2504. [Google Scholar] [CrossRef]
- Saini, A.; Singh, H. Enhanced human identity and gender recognition from gait sequences using SVM and MDA. Int. J. Comput. Appl. 2015, 119, 6–9. [Google Scholar] [CrossRef]
- Do, T.D.; Nguyen, V.H.; Kim, H. Real-time and robust multiple-view gender classification using gait features in video surveillance. Pattern Anal. Appl. 2020, 23, 399–413. [Google Scholar] [CrossRef]
- Li, H.; Qiu, Y.; Zhao, H.; Zhan, J.; Chen, R.; Wei, T.; Huang, Z. GaitSlice: A gait recognition model based on spatio-temporal slice features. Pattern Recognit. 2022, 124, 108453. [Google Scholar] [CrossRef]
- Gul, S.; Malik, M.I.; Khan, G.M.; Shafait, F. Multi-view gait recognition system using spatio-temporal features and deep learning. Expert Syst. Appl. 2021, 179, 115057. [Google Scholar] [CrossRef]
- Dong, D.; Ma, C.; Wang, M.; Vu, H.T.; Vanderborght, B.; Sun, Y. A low-cost framework for the recognition of human motion gait phases and patterns based on multi-source perception fusion. Eng. Appl. Artif. Intell. 2023, 120, 105886. [Google Scholar] [CrossRef]
- Wen, J.; Shen, Y.; Yang, J. Multi-view gait recognition based on generative adversarial network. Neural Process. Lett. 2022, 54, 1855–1877. [Google Scholar] [CrossRef]
- Zhang, Y.; Huang, Y.; Yu, S.; Wang, L. Cross-view gait recognition by discriminative feature learning. IEEE Trans. Image Process. 2020, 29, 1001–1015. [Google Scholar] [CrossRef]
- Luo, J.; Tjahjadi, T. Multi-set canonical correlation analysis for 3D abnormal gait behavior recognition based on virtual sample generation. IEEE Access 2020, 8, 32485–32501. [Google Scholar] [CrossRef]
- Shiraga, K.; Makihara, Y.; Muramatsu, D.; Echigo, T.; Yagi, Y. Geinet: View-invariant gait recognition using a convolutional neural network. In Proceedings of the International Conference on Biometrics, Halmstad, Sweden, 13–16 June 2016; pp. 1–8. [Google Scholar]
- Sakata, A.; Takemura, N.; Yagi, Y. Gait-based age estimation using multi-stage convolutional neural network. IPSJ Trans. Comput. Vis. Appl. 2019, 11, 4. [Google Scholar] [CrossRef]
- Xu, C.; Makihara, Y.; Liao, R.; Niitsuma, H.; Li, X.; Yagi, Y.; Lu, J. Real-time gait-based age estimation and gender classification from a single image. In Proceedings of the IEEE/CVF Conference on Applications of Computer Vision, Waikoloa, HI, USA, 3–8 January 2021; pp. 3460–3470. [Google Scholar]
- Xu, C.; Makihara, Y.; Li, X.; Yagi, Y.; Lu, J. Gait recognition from a single image using a phase-aware gait cycle reconstruction network. In Proceedings of the European Conference on Computer Vision, Glasgow, UK, 23–28 August 2020; pp. 386–403. [Google Scholar]
- Chao, H.; He, Y.; Zhang, J.; Feng, J. GaitSet: Regarding gait as a set for cross-view gait recognition. In Proceedings of the AAAI Conference on Artificial Intelligence, Honolulu, HI, USA, 27 January–1 February 2019; Volume 33, pp. 8126–8133. [Google Scholar]
- Kwon, B.; Lee, S. Joint swing energy for skeleton-based gender classification. IEEE Access 2021, 9, 28334–28348. [Google Scholar] [CrossRef]
- Zhao, L.; Guo, L.; Zhang, R.; Xie, X.; Ye, X. mmGaitSet: Multimodal based gait recognition for countering carrying and clothing changes. Appl. Intell. 2022, 52, 2023–2036. [Google Scholar] [CrossRef]
- Bei, S.; Deng, J.; Zhen, Z.; Shaojing, S. Gender recognition via fused silhouette features based on visual sensors. IEEE Sens. J. 2019, 19, 9496–9503. [Google Scholar] [CrossRef]
- Russel, N.S.; Selvaraj, A. Gender discrimination, age group classification and carried object recognition from gait energy image using fusion of parallel convolutional neural network. IET Image Process. 2021, 15, 239–251. [Google Scholar] [CrossRef]
- Fu, B.; Dong, Y.; Fu, S.; Wu, Y.; Ren, Y.; Thanh, D.N.H. Multistage supervised contrastive learning for hybrid-degraded image restoration. Signal Image Video Process. 2023, 17, 573–581. [Google Scholar] [CrossRef]
- Sethi, D.; Prakash, C.; Bharti, S. Multi-feature gait analysis approach using deep learning in constraint-free environment. Expert Syst. 2023, e13274. [Google Scholar] [CrossRef]
- Anbalagan, E.; Anbhazhagan, S.M. Deep learning model using ensemble based approach for walking activity recognition and gait event prediction with grey level co-occurrence matrix. Expert Syst. Appl. 2023, 227, 120337. [Google Scholar] [CrossRef]
- Mogan, J.N.; Lee, C.P.; Lim, K.M.; Ali, M.; Alqahtani, A. Gait-CNN-ViT: Multi-model gait recognition with convolutional neural networks and vision transformer. Sensors 2023, 23, 3809. [Google Scholar] [CrossRef]
- The OU-ISIR Gait Database, Large Population Dataset with Age. Available online: http://www.am.sanken.osaka-u.ac.jp/BiometricDB/GaitLPAge.html (accessed on 1 January 2023).
- The OU-ISIR Gait Database, Multi-View Large Population Dataset. Available online: http://www.am.sanken.osaka-u.ac.jp/BiometricDB/GaitMVLP.html (accessed on 1 January 2023).
- Zhang, D.; Wang, Y. Gender recognition based on fusion of face and gait information. In Proceeding of International Conference on Machine Learning and Cybernetics, Kunming, China, 12–15 July 2008; pp. 62–67. [Google Scholar]
- Samangooei, S.; Bustard, J.D.; Seely, R.D.; Nixon, M.S.; Carter, J.N. Acquisition and analysis of a dataset comprising gait, ear, and semantic data. In Multibiometrics for Human Identification; Cambridge University Press: Cambridge, UK, 2011; Chapter 12. [Google Scholar]
- CASIA Gait Database. Available online: http://www.sinobiometrics.com (accessed on 1 January 2023).
- Zhu, Z.; Guo, X.; Yang, T.; Huang, J.; Deng, J.; Huang, G.; Du, D.; Lu, J.; Zhou, J. Gait recognition in the wild: A benchmark. In Proceedings of the IEEE International Conference on Computer Vision, Montreal, BC, Canada, 11–17 October 2021; pp. 14789–14799. [Google Scholar]
- Isaac, E.R.; Elias, S.; Rajagopalan, S.; Easwarakumar, K. Multiview gait-based gender classification through pose-based voting. Pattern Recognit. Lett. 2019, 126, 41–50. [Google Scholar] [CrossRef]
- Głowiński, S.; Błażejewski, A.; Królikowski, T.; Knitter, R. Gait Recognition: A Challenging Task for MEMS Signal Identification. In Smart Innovation, Systems and Technologies; Springer: Singapore, 2019; Volume 155, Chapter 39; pp. 473–483. [Google Scholar]
Characteristics | CASIA B | OU-ISIR MVLP |
---|---|---|
Number of male subjects | 93 | 5114 |
Number of female subjects | 31 | 5193 |
Number of view angles | 11 | 14 |
View range | 0–180 | 0–90, 180–270 |
Age range | - | 2–87 |
Image resolution | 352 × 240 | 1280 × 980 |
Method | Recall | Precision | F1-score | Acc |
---|---|---|---|---|
FLBP* [13] | 0.902 | 0.907 | 0.904 | 0.903 |
PBV-EFD [41] | - | - | 0.953 | |
CNN C_customized [30] | 0.929 | 0.995 | 0.961 | 0.960 |
Our proposed model (Stage 1 + Stage 2B) | 0.982 | 0.979 | 0.980 | 0.981 |
Truth Predict | Female | Male | Total |
---|---|---|---|
Female | 23,028 | 1047 | 24,075 |
Male | 783 | 29,758 | 30,541 |
Total | 23,811 | 30,805 | 54,616 |
Walking Status | Female Recall | Female Precision | Female F1-score | Male Recall | Male Precision | Male F1-score | BA |
---|---|---|---|---|---|---|---|
Normal | 0.972 | 0.971 | 0.972 | 0.980 | 0.981 | 0.981 | 0.976 |
With a bag | 0.971 | 0.967 | 0.969 | 0.977 | 0.980 | 0.978 | 0.973 |
In a coat | 0.950 | 0.953 | 0.952 | 0.968 | 0.966 | 0.967 | 0.960 |
Walking Status | FLBP* [13] | Our Proposed Model | Difference |
---|---|---|---|
Normal | 0.964 | 0.963 | −0.001 |
With a bag | 0.880 | 0.961 | +0.081 |
In a coat | 0.865 | 0.941 | +0.076 |
Method | Acc |
---|---|
GEINet [22] | 0.939 |
GaitSet [26] | 0.927 |
Xu [24] | 0.943 |
Our proposed model (Stage 1 + Stage 2A) | 0.948 |
Truth Predict | Female | Male | Total |
---|---|---|---|
Female | 28,441 | 1730 | 30,171 |
Male | 1336 | 27,923 | 29,259 |
Total | 29,777 | 29,653 | 59,430 |
Age Group | Female Recall | Female Precision | Female F1-Score | Male Recall | Male Precision | Male F1-Score | BA |
---|---|---|---|---|---|---|---|
0–5 | 0.753 | 0.913 | 0.825 | 0.909 | 0.742 | 0.817 | 0.831 |
6–10 | 0.791 | 0.848 | 0.819 | 0.857 | 0.803 | 0.829 | 0.824 |
11–15 | 0.864 | 0.853 | 0.858 | 0.863 | 0.874 | 0.868 | 0.864 |
16–60 | 0.927 | 0.856 | 0.890 | 0.858 | 0.928 | 0.892 | 0.893 |
Over 60 | 0.893 | 0.640 | 0.746 | 0.681 | 0.909 | 0.779 | 0.787 |
Model | Female Recall | Female Precision | Female F1-Score | Male Recall | Male Precision | Male F1-Score | Acc |
---|---|---|---|---|---|---|---|
Single-stream Stage 1 (only GEI CNN) | 0.870 | 0.881 | 0.875 | 0.879 | 0.869 | 0.874 | 0.875 |
Three-stream Stage 1 + Stage 2A | 0.970 | 0.980 | 0.975 | 0.984 | 0.976 | 0.980 | 0.978 |
Three-stream Stage 1 + Stage 2B | 0.992 | 0.964 | 0.978 | 0.973 | 0.994 | 0.983 | 0.981 |
Three-stream Stage 1 + Stage 2C | 0.977 | 0.964 | 0.970 | 0.972 | 0.982 | 0.977 | 0.974 |
Model | Female Recall | Female Precision | Female F1-Score | Male Recall | Male Precision | Male F1-Score | Acc |
---|---|---|---|---|---|---|---|
Single-stream Stage 1 (only GEI CNN) | 0.844 | 0.778 | 0.810 | 0.789 | 0.852 | 0.819 | 0.815 |
Three-stream Stage 1 + Stage 2A | 0.955 | 0.943 | 0.949 | 0.942 | 0.954 | 0.948 | 0.948 |
Three-stream Stage 1 + Stage 2B | 0.942 | 0.947 | 0.944 | 0.945 | 0.940 | 0.942 | 0.943 |
Three-stream Stage 1 + Stage 2C | 0.944 | 0.946 | 0.945 | 0.944 | 0.942 | 0.943 | 0.944 |
Dropout Rate | AUC |
---|---|
0.1 | 0.961 |
0.2 | 0.961 |
0.3 | 0.961 |
0.4 | 0.965 |
0.5 | 0.967 |
0.6 | 0.953 |
0.7 | 0.966 |
Learning Rate | Stage 1 + Stage 2A | Stage 1 + Stage 2B | Stage 1 + Stage 2C |
---|---|---|---|
0.0008 | 0.976 | 0.965 | 0.961 |
0.0009 | 0.970 | 0.970 | 0.955 |
0.0010 | 0.982 | 0.972 | 0.974 |
0.0011 | 0.978 | 0.957 | 0.971 |
0.0012 | 0.977 | 0.968 | 0.969 |
Model | Inference Time (s) | Number of Parameters |
---|---|---|
Stage 1 | 0.029 | 52 M |
Stage 1 + Stage 2A | 0.305 | 62 M |
Stage 1 + Stage 2B | 0.327 | 53 M |
Stage 1 + Stage 2C | 0.308 | 53 M |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Leung, T.-M.; Chan, K.-L. Gender Recognition Based on Gradual and Ensemble Learning from Multi-View Gait Energy Images and Poses. Sensors 2023, 23, 8961. https://doi.org/10.3390/s23218961
Leung T-M, Chan K-L. Gender Recognition Based on Gradual and Ensemble Learning from Multi-View Gait Energy Images and Poses. Sensors. 2023; 23(21):8961. https://doi.org/10.3390/s23218961
Chicago/Turabian StyleLeung, Tak-Man, and Kwok-Leung Chan. 2023. "Gender Recognition Based on Gradual and Ensemble Learning from Multi-View Gait Energy Images and Poses" Sensors 23, no. 21: 8961. https://doi.org/10.3390/s23218961
APA StyleLeung, T.-M., & Chan, K.-L. (2023). Gender Recognition Based on Gradual and Ensemble Learning from Multi-View Gait Energy Images and Poses. Sensors, 23(21), 8961. https://doi.org/10.3390/s23218961