A Resource Constrained Neural Network for the Design of Embedded Human Posture Recognition Systems
Abstract
:1. Introduction
- the capability of the FCN to achieve high recognition accuracy by monitoring only the footprint of the human body in a limited space region covered by a reduced number of pressure sensors.
- Any sensor placement strategy is unnecessary, namely the system reliability is not dependent on the specific support.
- FCN can be easily reconfigured to different applications. Case studies are presented on laying and sitting postures recognition.
- FCN provides end-to-end classification by using a quantization scheme that overcomes binarized and ternary counterparts in terms of accuracy and meets the optimal trade-off between accuracy and employed physical resources for HW implementation.
2. The Proposed System and the Underlying Model
2.1. The FCN Model
2.2. FCN Training and Accuracy Results
3. System Design
4. Synthesis and Implementation Results
5. Comparison with the Literature
6. Conclusions
Author Contributions
Funding
Conflicts of Interest
References
- Boulay, B.; Brémond, F.; Thonnat, M. Applying 3D human model in a posture recognition system. Pattern Recognit. Lett. 2006, 27, 1788–1796. [Google Scholar] [CrossRef] [Green Version]
- Ni, W.; Gao, Y.; Lucev, Z.; Pun, S.H.; Cifrek, M.; Vai, M.I.; Du, M. Human posture detection based on human body communication with muti-carriers modulation. In Proceedings of the 2016 39th International Convention on Information and Communication Technology, Electronics and Microelectronics (MIPRO), Opatija, Croatia, 30 May–3 June 2016; pp. 273–276. [Google Scholar]
- De Vita, A.; Pau, D.; Di Benedetto, L.; Rubino, A.; Pétrot, F.; Licciardo, G.D. Low Power Tiny Binary Neural Network with improved accuracy in Human Recognition Systems. In Proceedings of the 2020 23rd Euromicro Conference on Digital System Design (DSD), Kranj, Slovenia, 26–28 August 2020; pp. 309–315. [Google Scholar]
- De Vita, A.; Russo, A.; Pau, D.; Di Benedetto, L.; Rubino, A.; Licciardo, G.D. A Partially Binarized Hybrid Neural Network System for Low-Power and Resource Constrained Human Activity Recognition. IEEE Trans. Circuits Syst. I Regul. Pap. 2020, 67, 3893–3904. [Google Scholar] [CrossRef]
- De Vita, A.; Pau, D.; Parrella, C.; Di Benedetto, L.; Rubino, A.; Licciardo, G.D. Low-Power HWAccelerator for AI Edge-Computing in Human Activity Recognition Systems. In Proceedings of the 2020 2nd IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS), Genova, Italy, 31 August–4 September 2020; pp. 291–295. [Google Scholar]
- Tattersall, R.; Walshaw, M. Posture and cystic fibrosis. J. R. Soc. Med. 2003, 96, 18. [Google Scholar]
- Grandjean, E.; Hünting, W. Ergonomics of posture—review of various problems of standing and sitting posture. Appl. Ergon. 1977, 8, 135–140. [Google Scholar] [CrossRef]
- O’Sullivan, K.; O’Dea, P.; Dankaerts, W.; O’Sullivan, P.; Clifford, A.; O’Sullivan, L. Neutral lumbar spine sitting posture in pain-free subjects. Man. Ther. 2010, 15, 557–561. [Google Scholar] [CrossRef] [PubMed] [Green Version]
- Grujicic, M.; Pandurangan, B.; Xie, X.; Gramopadhye, A.; Wagner, D.; Ozen, M. Musculoskeletal computational analysis of the influence of car-seat design/adjustments on long-distance driving fatigue. Int. J. Ind. Ergon. 2010, 40, 345–355. [Google Scholar] [CrossRef]
- Fiorillo, I.; Piro, S.; Anjani, S.; Smulders, M.; Song, Y.; Naddeo, A.; Vink, P. Future vehicles: The effect of seat configuration on posture and quality of conversation. Ergonomics 2019, 62, 1400–1414. [Google Scholar] [CrossRef]
- Cardoso, M.; McKinnon, C.; Viggiani, D.; Johnson, M.J.; Callaghan, J.P.; Albert, W.J. Biomechanical investigation of prolonged driving in an ergonomically designed truck seat prototype. Ergonomics 2018, 61, 367–380. [Google Scholar] [CrossRef]
- Fasulo, L.; Naddeo, A.; Cappetti, N. A study of classroom seat (dis) comfort: Relationships between body movements, center of pressure on the seat, and lower limbs’ sensations. Appl. Ergon. 2019, 74, 233–240. [Google Scholar] [CrossRef]
- Hiemstra-van Mastrigt, S.; Groenesteijn, L.; Vink, P.; Kuijt-Evers, L.F. Predicting passenger seat comfort and discomfort on the basis of human, context and seat characteristics: A literature review. Ergonomics 2017, 60, 889–911. [Google Scholar] [CrossRef] [PubMed]
- Wafaa, M.; Salih Abedi, E.A. Modified Deep Learning Method for Body Postures Recognition. Int. J. Adv. Sci. Technol. 2020, 29, 3830–3841. [Google Scholar]
- Han, S.H.; Kim, H.G.; Choi, H.J. Rehabilitation posture correction using deep neural network. In Proceedings of the 2017 IEEE international conference on big data and smart computing (BigComp), Jeju Island, Korea, 13–16 February 2017; pp. 400–402. [Google Scholar]
- Elforaici, M.E.A.; Chaaraoui, I.; Bouachir, W.; Ouakrim, Y.; Mezghani, N. Posture recognition using an RGB-D camera: Exploring 3D body modeling and deep learning approaches. In Proceedings of the 2018 IEEE life sciences conference (LSC), Montreal, QC, Canada, 28–30 October 2018; pp. 69–72. [Google Scholar]
- Matar, G.; Lina, J.M.; Kaddoum, G. Artificial neural network for in-bed posture classification using bed-sheet pressure sensors. IEEE J. Biomed. Health Inform. 2019, 24, 101–110. [Google Scholar] [CrossRef] [PubMed]
- Ren, W.; Ma, O.; Ji, H.; Liu, X. Human Posture Recognition Using a Hybrid of Fuzzy Logic and Machine Learning Approaches. IEEE Access 2020, 8, 135628–135639. [Google Scholar] [CrossRef]
- Lin, G.; Zhan, Z.; Peng, X.; Xu, H.; Fu, Y.; Jiang, L. A Study of Driver’s Driving Concentration Based on Computer Vision Technology; SAE Technical Paper; SAE International: Warrendale, PA, USA, 2020. [Google Scholar]
- Wachs, J.P.; Kölsch, M.; Goshorn, D. Human posture recognition for intelligent vehicles. J. Real Time Image Process. 2010, 5, 231–244. [Google Scholar] [CrossRef]
- Loeb, H.; Kim, J.; Arbogast, K.; Kuo, J.; Koppel, S.; Cross, S.; Charlton, J. Automated recognition of rear seat occupants’ head position using Kinect™ 3D point cloud. J. Saf. Res. 2017, 63, 135–143. [Google Scholar] [CrossRef] [PubMed]
- Zhao, Z.; Xia, S.; Xu, X.; Zhang, L.; Yan, H.; Xu, Y.; Zhang, Z. Driver Distraction Detection Method Based on Continuous Head Pose Estimation. Comput. Intell. Neurosci. 2020, 2020, 9606908. [Google Scholar] [CrossRef]
- Tlili, F.; Haddad, R.; Ouakrim, Y.; Bouallegue, R.; Mezghani, N. A Review on posture monitoring systems. In Proceedings of the 2018 International Conference on Smart Communications and Networking (SmartNets), Yasmine Hammamet, Tunisia, 16–17 November 2018; pp. 1–6. [Google Scholar]
- Viriyavit, W.; Sornlertlamvanich, V.; Kongprawechnon, W.; Pongpaibool, P.; Isshiki, T. Neural network based bed posture classification enhanced by Bayesian approach. In Proceedings of the 2017 8th International Conference of Information and Communication Technology for Embedded Systems (IC-ICTES), Chonburi, Thailand, 7–9 May 2017; pp. 1–5. [Google Scholar]
- Wang, J.; Hafidh, B.; Dong, H.; El Saddik, A. Sitting Posture Recognition Using a Spiking Neural Network. IEEE Sens. J. 2020, 21, 1779–1786. [Google Scholar] [CrossRef]
- Cappetti, N.; Di Manso, E. Study of the relationships between articular moments, comfort and human posture on a chair. Work A J. Prev. Assess. Rehabil. 2021, 68, S59–S68. [Google Scholar]
- Hu, Q.; Tang, X.; Tang, W. A smart chair sitting posture recognition system using flex sensors and FPGA implemented artificial neural network. IEEE Sens. J. 2020, 20, 8007–8016. [Google Scholar] [CrossRef]
- Bianchi, V.; Bassoli, M.; Lombardo, G.; Fornacciari, P.; Mordonini, M.; De Munari, I. IoT wearable sensor and deep learning: An integrated approach for personalized human activity recognition in a smart home environment. IEEE Internet Things J. 2019, 6, 8553–8562. [Google Scholar] [CrossRef]
- Rahimiazghadi, M.; Lammie, C.; Eshraghian, J.K.; Payvand, M.; Donati, E.; Linares-Barranco, B.; Indiveri, G. Hardware implementation of deep network accelerators towards healthcare and biomedical applications. IEEE Trans. Biomed. Circuits Syst. 2020, 6, 1138–1159. [Google Scholar] [CrossRef] [PubMed]
- Long, X.; Zeng, X.; Ben, Z.; Zhou, D.; Zhang, M. A Novel Low-Bit Quantization Strategy for Compressing Deep Neural Networks. Comput. Intell. Neurosci. 2020, 2020, 7839064. [Google Scholar] [CrossRef] [PubMed]
- Courbariaux, M.; Hubara, I.; Soudry, D.; El-Yaniv, R.; Bengio, Y. Binarized neural networks: Training deep neural networks with weights and activations constrained to+ 1 or-1. arXiv 2016, arXiv:1602.02830. [Google Scholar]
- Licciardo, G.D.; Cappetta, C.; Di Benedetto, L.; Rubino, A.; Liguori, R. Multiplier-less stream processor for 2D filtering in visual search applications. IEEE Trans. Circuits Syst. Video Technol. 2016, 28, 267–272. [Google Scholar] [CrossRef]
- Licciardo, G.D.; Cappetta, C.; Di Benedetto, L.; Vigliar, M. Weighted partitioning for fast multiplierless multiple-constant convolution circuit. IEEE Trans. Circuits Syst. II Express Briefs 2016, 64, 66–70. [Google Scholar] [CrossRef]
- Pouyan, M.B.; Birjandtalab, J.; Heydarzadeh, M.; Nourani, M.; Ostadabbas, S. A pressure map dataset for posture and subject analytics. In Proceedings of the 2017 IEEE EMBS International Conference on Biomedical & Health Informatics (BHI), Orlando, FL, USA, 16–19 February 2017; pp. 65–68. [Google Scholar]
- Zhou, B.; Khosla, A.; Lapedriza, A.; Oliva, A.; Torralba, A. Learning deep features for discriminative localization. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA, 27–30 June 2016; pp. 2921–2929. [Google Scholar]
- Diao, H.; Chen, C.; Yuan, W.; Amara, A.; Tamura, T.; Fan, J.; Meng, L.; Liu, X.; Chen, W. Deep Residual Networks for Sleep Posture Recognition With Unobtrusive Miniature Scale Smart Mat System. IEEE Trans. Biomed. Circuits Syst. 2021, 15, 111–121. [Google Scholar] [CrossRef] [PubMed]
Layer | N° Parameters | Operations per Windows | N° of Bits |
---|---|---|---|
CONV 1 | SAC: (* | ||
NORM 1 | ADD: (* | ||
CONV 2 | SAC: (* | ||
NORM 2 | ADD: (* | ||
CONV 3 | SAC: (* | ||
NORM 3 | ADD: (* | ||
GAP | 0 | ADD: (* | 0 |
SOFTMAX | Classes ** | SAC: Classes ** |
Dataset | PMatData:Laying Posture | Custom:Sitting Posture |
---|---|---|
Number of classes | 17 | 8 |
Available classes | Supine (9 types), right, | Initial position, bent forward, |
right (30°),right (60°), | rested back, bent left, legs up, | |
right fetus, left, left (30°), | right-bent thinker, straight legs, | |
left (60°), left fetus. | left-bent thinker | |
Training Set | 15,232 | 6720 |
Test set | 1692 | 1680 |
Class | Precision | Recall | F1-Score | Support |
---|---|---|---|---|
Supine | 100 | |||
Right | 0.92 | 1.00 | 0.96 | 102 |
Left | 0.96 | 0.97 | 0.96 | 96 |
Right 30° (1 wedge) | 0.99 | 1.00 | 1.00 | 114 |
Right 60° (2 wedges) | 1.00 | 1.00 | 1.00 | 96 |
Left 30° (1 wedge) | 1.00 | 0.95 | 0.97 | 99 |
Left 60° (2 wedges) | 0.97 | 0.98 | 0.97 | 88 |
Supine 1 | 0.94 | 1.00 | 0.97 | 102 |
Supine 2 | 0.98 | 0.99 | 0.99 | 120 |
Supine 3 | 0.99 | 1.00 | 0.99 | 88 |
Supine 4 | 0.99 | 0.98 | 0.99 | 117 |
Supine 5 | 1.00 | 0.99 | 0.99 | 98 |
Right Fetus | 1.00 | 0.96 | 0.98 | 90 |
Left Fetus | 1.00 | 0.98 | 0.99 | 99 |
Supine (30°) | 0.97 | 0.97 | 0.97 | 101 |
Supine (45°) | 0.97 | 0.89 | 0.93 | 88 |
Supine (60°) | 0.94 | 1.00 | 0.97 | 94 |
Class | Precision | Recall | F1-Score | Support |
---|---|---|---|---|
Initial position | 0.99 | 0.99 | 0.99 | 839 |
Bent forward | 0.99 | 0.98 | 0.98 | 128 |
Rested back | 1 | 0.98 | 0.99 | 123 |
Bent left | 0.98 | 1 | 0.99 | 125 |
Left-bent thinker | 1 | 1 | 1 | 116 |
Right-bent thinker | 1 | 0.98 | 0.99 | 114 |
Legs up | 0.98 | 0.99 | 0.99 | 120 |
Straight legs | 1 | 1 | 1 | 115 |
Proposed Sitting | Proposed Laying | |
---|---|---|
Neural Network | FCN | FCN |
Network Complexity | 3 CONV+GAP+FC | 3 CONV+GAP+FC |
N° Classes | 8 | 17 |
Mean Accuracy | 98.81% | 96.77% |
Target Platform | Artix-7 | Artix-7 |
N° sensors | 56 | 108 |
Sensing area [mm] | 300 × 200 carpet | 320 × 200 carpet |
Dynamic Power [/] | 144 | 391 |
Dyn.Power @MaxFreq [] | 10.4 | 6.88 |
Tot. Power @OpFreq [] | 72 | 72 |
N° LUTs | 15,802 | 10,983 |
N° FFs | 12,287 | 8424 |
N° DSPs | 0 | 0 |
N° BRAMs | 0 | 0 |
Max Freq. [] | 47.64 | 26.6 |
Max Sensor ODR [] | 16.50 | 9.13 |
Delay @ Max Freq [] | 60 | 109 |
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2021 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
Licciardo, G.D.; Russo, A.; Naddeo, A.; Cappetti, N.; Di Benedetto, L.; Rubino, A.; Liguori, R. A Resource Constrained Neural Network for the Design of Embedded Human Posture Recognition Systems. Appl. Sci. 2021, 11, 4752. https://doi.org/10.3390/app11114752
Licciardo GD, Russo A, Naddeo A, Cappetti N, Di Benedetto L, Rubino A, Liguori R. A Resource Constrained Neural Network for the Design of Embedded Human Posture Recognition Systems. Applied Sciences. 2021; 11(11):4752. https://doi.org/10.3390/app11114752
Chicago/Turabian StyleLicciardo, Gian Domenico, Alessandro Russo, Alessandro Naddeo, Nicola Cappetti, Luigi Di Benedetto, Alfredo Rubino, and Rosalba Liguori. 2021. "A Resource Constrained Neural Network for the Design of Embedded Human Posture Recognition Systems" Applied Sciences 11, no. 11: 4752. https://doi.org/10.3390/app11114752