面向外骨骼性能评估的多源生理能耗预测

doi:10.12068/j.issn.1005-3026.2024.06.013

摘要/Abstract

摘要：

在外骨骼设计过程中，助力性能评估直接影响整体结构的安全性和效率.针对目前性能评估多采用单一指标问题，提出了一种基于多源生理信号（表面肌电、脉搏和呼吸）的长短期记忆（long short?term memory，LSTM）网络运动能耗预测方法.该方法首先对生理信号进行处理和特征提取，然后使用6层LSTM模型进行预测，并使用K折交叉验证方法进行验证.同时，与决策树（decision tree，DT）和支持向量机（support vector machine，SVM）进行了对比实验.最后，建立了在线能耗监测系统，作为评估外骨骼助力性能的依据.结果表明，三源信号均方根差（root mean square error，RMSE）为0.073 kJ，证明了采用多源生理信号融合预测的可行性；LSTM的RMSE相较于DT，SVM下降了39.53%，15.68%，测试集的总能耗值误差值为23.98 kJ，说明LSTM模型对于总能耗预测效果最好，可用于外骨骼的助力性能评估.

关键词: 能耗预测, 长短期记忆, 外骨骼性能评估, 信号融合, 在线监测

Abstract:

In the process of exoskeleton design， the evaluation of assistance performance directly impacts the overall structural safety and efficiency. Addressing the current issue of predominantly utilizing single metrics for performance evaluation， a method based on multi?source physiological signals （surface electromyography， photopretismography， and respiration） for LSTM prediction of motion energy consumption was proposed. This method involves preprocessing and feature extraction of physiological signals， followed by prediction using a 6?layer LSTM model and validation through K-fold cross?validation. Comparative experiments with DT and SVM were conducted. Finally， an online energy consumption monitoring system was established， serving as a basis for evaluating exoskeleton assistance performance. Results indicate the feasibility of utilizing multi?source physiological signals for fusion prediction， with an RMSE of 0.073 kJ for the three?source signal. The LSTM model achieves a 39.53% and 15.68% reduction in RMSE compared to DT and SVM， respectively. The total energy consumption error on the test set is 23.98 kJ， demonstrating the superior performance of the LSTM model for total energy consumption prediction and its suitability for exoskeleton assistance performance evaluation.

Key words: energy consumption prediction, LSTM, exoskeleton performance evaluation, signal fusion, online monitoring

中图分类号:

TP 183

李坦, 王宏, 金博丕, 吴志伟. 面向外骨骼性能评估的多源生理能耗预测[J]. 东北大学学报（自然科学版）, 2024, 45(6): 850-857.

Tan LI, Hong WANG, Bo-pi JIN, Zhi-wei WU. Multi-source Physiological Energy Consumption Prediction for Exoskeleton Performance Evaluation[J]. Journal of Northeastern University(Natural Science), 2024, 45(6): 850-857.

图/表 13

图1 信号采集传感器穿戴示意图

Fig.1 Schematic diagram of the signal acquisition sensor worn

图2 三种信号时域图（a）—sEMG信号；（b）—PPG信号；（c）—RESP信号.

Fig.2 Time?domain diagrams of the three signals

图3 时间窗口移动及每名受试者数据集构成示意图

Fig.3 Diagram of time window movement and data set composition for each subject

图4 LSTM网络整体结构

Fig.4 Overall structure of the LSTM network

图5 LSTM模型结构图

Fig.5 LSTM model structure diagram

图6 运动能耗预测流程图

Fig.6 Flow chart for predicting energy consumption in motion

表1 不同特征组合交叉验证

Table 1 Cross?validation of different feature combinations

组合类型	信号名称	特征个数	RMSE/kJ
单一源信号	s	5	0.133
	P	4	0.116
	R	3	0.120
双源信号	s-P	8	0.089
	s-R	7	0.087
	P-R	6	0.080
三源信号	s-P-R	10	0.073

表2 预测模型评价指标对比 (indicators kJ)

Table 2 Comparison of predictive model evaluation

方法	RMSE均值	RMSE方差
DT	0.108	0.009 0
SVM	0.077	0.008 0
LSTM	0.065	0.004 5

图7 DT，SVM和LSTM模型预测效果图

Fig.7 Prediction results of DT，SVM and LSTM models

图8 测试集数据预测的总能耗值对比

Fig.8 Comparison of total energy consumption values predicted from test set data

图9 柔性外骨骼穿戴效果图

Fig.9 Wearing effect of the flexible exoskeleton

图10 基于LabVIEW的能耗在线监测系统界面

Fig.10 Interface of online energy consumption monitoring system based on LabVIEW

图11 外骨骼能耗评估流程图

Fig.11 Flow chart for exoskeleton energy consumption evaluation

参考文献 14

1	Shepertycky M， Burton S， Dickson A，et al.Removing energy with an exoskeleton reduces the metabolic cost of walking［J］.Science，2021，372（6545）：957-960.
2	戚文谦，孙守迁，陈超，等.面向竞技训练的下肢外骨骼多模态交互系统研究［J］.机械工程学报，2023，59（5）：29-40.
	Qi Wen‑qian， Sun Shou‑qian， Chen Chao，et al.Research on multimodal interaction system of lower limb exoskeleton for athletic training［J］.Journal of Mechanical Engineering，2023，59（5）：29-40.
3	Baltrusch S J， van Dieën J H， Koopman A S，et al.SPEXOR passive spinal exoskeleton decreases metabolic cost during symmetric repetitive lifting［J］.European Journal of Applied Physiology，2020，120（2）：401-412.
4	Zhou T C， Xiong C H， Zhang J J，et al.Reducing the metabolic energy of walking and running using an unpowered hip exoskeleton［J］.Journal of Neuro Engineering and Rehabilitation，2021，18（1）：1-15.
5	Slade P， Kochenderfer M J， Delp S L，et al.Personalizing exoskeleton assistance while walking in the real world［J］.Nature，2022，610（7931）：277-282.
6	Beato M， Impellizzeri F M， Coratella G，et al.Quantification of energy expenditure of recreational football［J］.Journal of Sports Sciences，2016，34（24）：2185-2188.
7	Bisi M C， Stagni R， Han H，et al.An EMG‑driven model applied for predicting metabolic energy consumption during movement［J］.Journal of Electromyography & Kinesiology，2011，21（6）：1074-1080.
8	Tsianos G A， Macfadden L N.Validated predictions of metabolic energy consumption for submaximal effort movement［J］.Plos Computational Biology，2016，12（6）：e1004911.
9	Choi J W， Kim H J.Accuracy verification of heart rate and energy consumption tracking devices to develop forest‑based customized health care service programs［J］.Journal of People Plants and Environment，2019，22（2）：219-229.
10	Wang B C， Li Y， Luo Y，et al.Early event detection in a deep‑learning driven quality prediction model for ultrasonic welding［J］.Journal of Manufacturing Systems，2021，60：325-336.
11	王斐，魏晓童，秦皞.基于sEMG和LSTM的下肢连续运动估计［J］.东北大学学报（自然科学版），2020，41（3）：305-310.
	Wang Fei， Wei Xiao‑tong， Qin Hao.Continuous motion estimation of lower limbs based on sEMG and LSTM［J］.Journal of Northeastern University（Natural Science），2020，41（3）：305-310.
12	仇逊超，曹军.无源无线传感器平台在人体跑步检测中的应用［J］.计算机工程与设计，2014，35（12）：4395-4401.
	Qiu Xun‑chao， Cao Jun.Application of passive wireless sensor platform in human running detection［J］.Computer Engineering and Design，2014，35（12）：4395-4401.
13	Zong X， Wang F， She J，et al.Denoising method for surface electromyography signals combining CEEMDAN and interval total variation［J］.Circuits，Systems，and Signal Processing，2022，41（11）：6493-6512.
14	Hochreiter S， Schmidhuber J.Long short‑term memory［J］.Neural Computation，1997，9（8）：1735-1780.

[1]	韩东红, 孔彦茹, 展艺萌, 刘源. 音乐多模态数据情感识别方法的研究[J]. 东北大学学报（自然科学版）, 2024, 45(6): 776-785.
[2]	杨譞，何占奇. 改进的两层BiLSTM的心电信号分割方法[J]. 东北大学学报（自然科学版）, 2023, 44(12): 1705-1711.
[3]	季策，王鑫，耿蓉，梁敏骏. 时变信道下基于LSTM的信道估计方法[J]. 东北大学学报（自然科学版）, 2023, 44(11): 1521-1528.
[4]	刘伟嵬，邓剑洋，张靖文，牛东东. 基于深度学习的挖掘机工作阶段的分类与识别[J]. 东北大学学报（自然科学版）, 2023, 44(10): 1464-1474.
[5]	赵骧;蒋奇武;左良;梁志德. 体心立方单晶体塑性应变比的计算[J]. 东北大学学报:自然科学版, 2002, 23(2): 134-136.
[6]	刘贵杰;巩亚东;王宛山. 基于神经网络的磨削砂轮状态的在线监测[J]. 东北大学学报:自然科学版, 2002, 23(10): 984-987.