Menu
生理约束肌肉骨骼神经网络（MSK-NN）从部分观测sEMG估计多自由度关节运动学，无需内部生物力学标签的直接监督。
The review-first prompt is selected by default. You can switch to a direct command or download a local copy.
Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
npx skills add https://github.com/hiyenwong/ai_collection --skill physiologically-constrained-musculoskeletal-neural-networkThe command stays on one line. Scroll horizontally to inspect it before copying.
Prefer a local copy? Download the files currently available to SkillsMP.
Graph-native Python reimplementation of the Information Dynamics of Music (IDyOM) model that represents predictive memories as explicit graph objects for musical expectation modeling and network analysis.
Physics-aware end-to-end deep reinforcement learning methodology for quadcopter control with actuator dynamics modeling.
Reinforced Dreamer methodology for asymmetric reinforcement learning using latent guidance to improve world model representations and behaviors in model-based RL.
Based on SOC occupation classification
Showing SKILL.md
| name | physiologically-constrained-musculoskeletal-neural-network |
| description | 生理约束肌肉骨骼神经网络（MSK-NN）从部分观测sEMG估计多自由度关节运动学，无需内部生物力学标签的直接监督。 |
| triggers | ["sEMG","表面肌电图","肌肉骨骼神经网络","MSK-NN","关节运动学估计","多自由度","multi-DoF","生理约束","physiologically constrained","肌肉激活估计","muscle activation","肌肉协同","muscle synergy","生物力学建模","biomechanical modeling"] |
| version | 1 |
| author | Wending Heng, Mingming Zhang, Glen Cooper, Zhenhong Li (arXiv:2606.07476v1) |
| arxiv_id | 2606.07476v1 |
| date_imported | 2026-06-08T00:00:00.000Z |
| source_url | https://arxiv.org/abs/2606.07476v1 |
| pdf_url | https://arxiv.org/pdf/2606.07476v1.pdf |
| tags | ["electromyography","neural network","biomechanics","joint kinematics","musculoskeletal modeling","physics-informed neural networks","physiological constraints","muscle synergy","rehabilitation","prosthetics"] |
提出新型肌肉骨骼神经网络（MSK-NN），从部分观测的表面肌电图（sEMG）估计多自由度关节角度，同时推断测量和未测量肌肉的激活。核心创新是无需内部生物力学变量（肌肉肌腱力、关节力矩）的直接监督。
现实约束：
现有方法局限：
CNN肌肉激活估计器：
嵌入MSK前向动力学模块：
关键区别：
| 方法 | 内部标签 | MSK-NN |
|---|---|---|
| 混合NN | 需肌肉力 | 无 |
| 传统NN | 需力矩 | 无 |
| MSK-NN | 无 | ✓ |
三重约束设计：
L_total = L_kinematics + L_synergy + L_trend
组成部分：
L_kinematics = MSE(θ_pred, θ_true)
确保关节角度估计准确
L_synergy = ||A - WS||
数据驱动肌肉协同约束
L_trend = ∑_i ||∂a_i/∂θ - ∂a_i_ref/∂θ||
解剖学引导的趋势约束
参数范围：
优化过程：
θ_MSK ∈ [θ_min, θ_max]
参数保持在生理极限内
双任务：
验证方法：
动力学方程嵌入：
τ_joint = ∑_i f_i(θ_MSK, a_i) × r_i(θ)
力矩计算物理约束
运动类型：
对比方法：
| 方法 | NRMSE | R² |
|---|---|---|
| CNN | 0.12 | 0.85 |
| Bi-LSTM | 0.11 | 0.87 |
| CNN-LSTM | 0.10 | 0.89 |
| PET | 0.09 | 0.91 |
| MSK-NN | 0.07 | 0.94 |
特别在随机运动：
参数验证：
意义：
突破：
优势：
class MuscleActivationEstimator(nn.Module):
def __init__(self, input_channels, n_muscles):
super().__init__()
# 时序特征提取
self.temporal_conv = nn.Conv1d(input_channels, 64, kernel_size=3)
# 空间特征融合
self.spatial_conv = nn.Conv1d(64, 128, kernel_size=1)
# 肌肉激活输出
self.activation_head = nn.Linear(128, n_muscles)
def forward(self, sEMG):
# 特征提取
temporal_features = F.relu(self.temporal_conv(sEMG))
spatial_features = F.relu(self.spatial_conv(temporal_features))
# 肌肉激活
activations = self.activation_head(spatial_features)
return activations
class MSKDynamicsModule(nn.Module):
def __init__(self, n_muscles, n_joints, MSK_params):
super().__init__()
self.muscle_params = MSK_params
self.n_joints = n_joints
def forward(self, activations, joint_angles):
"""
肌肉骨骼前向动力学
Args:
activations: 肌肉激活 [n_muscles]
joint_angles: 关节角度 [n_joints]
Returns:
joint_angles_est: 估计关节角度
activations_est: 估计肌肉激活
"""
# 肌肉力计算
muscle_forces = compute_muscle_force(
activations,
self.muscle_params,
joint_angles
)
# 力臂计算
moment_arms = compute_moment_arms(
joint_angles,
self.muscle_params
)
# 关节力矩
joint_torques = torch.sum(muscle_forces * moment_arms, dim=0)
# 关节角度估计
joint_angles_est = forward_dynamics(joint_torques)
return joint_angles_est, activations
def compute_total_loss(joint_pred, joint_true,
activations, synergy_weights):
# 关节运动学损失
L_kinematics = F.mse_loss(joint_pred, joint_true)
# 肌肉协同损失
synergy_matrix = activations @ synergy_weights.T
L_synergy = torch.norm(activations - synergy_matrix)
# 解剖趋势损失
d_activation = torch.autograd.grad(
activations.sum(), joint_pred, retain_graph=True
)[0]
L_trend = torch.norm(d_activation - reference_trend)
# 总损失
L_total = L_kinematics + λ1*L_synergy + λ2*L_trend
return L_total
用途：
应用：
诊断：
支持：
准确性：
揭示：
传统要求：
MSK-NN优势：
局限性：
MSK-NN改进：
创新： 首次实现无生物力学内部变量监督的肌肉骨骼神经网络训练
突破： 从部分肌肉数据推断完整肌肉激活
原创： 复合物理-生理损失函数创新设计
覆盖：
创新实验：
验证策略：
扩展方向：
深化研究：
工程实现：
Activation: sEMG处理、关节运动学估计、肌肉骨骼建模、生理约束神经网络、肌肉激活推断、肌肉协同分析、生物力学估计、康复监测、义肢控制