摘要
文章提出了一种基于机器学习的膝关节损伤检测方法。该方法利用加速度计采集的膝关节摆动信号,首先通过小波变换降低信号中的噪声能量,从而提高信噪比。接着,利用小波包分解提取小波能量,并通过梅林滤波器组计算信号的梅林倒谱系数。随后,将小波能量与梅林倒谱系数融合,形成融合特征,并通过主成分分析去除冗余信息。最后,采用最小二乘支持向量机、径向基神经网络和贝叶斯网络对健康和受损的膝关节摆动信号进行分类。实验结果表明,与现有方法相比,该方法在膝关节损伤检测方面具有更高的准确率。
This paper proposes a machine learning-based knee injury detection method.The method utilises the knee oscillation signal acquired by an accelerometer,and first reduces the noise energy in the signal by wavelet transform,thus improving the signal-to-noise ratio.Next,the wavelet energy is extracted using wavelet packet decomposition and the Merlin cepstrum coefficients of the signal are calculated by the Merlin filter bank.Subsequently,the wavelet energy is fused with the Merlin cepstrum coefficients to form a fusion feature,and the redundant information is removed by principal component analysis.Finally,least squares support vector machine,radial basis neural network and Bayesian network were used to classify healthy and damaged knee oscillating signals.The experimental results show that the method has higher accuracy in knee injury detection compared with existing methods.
作者
朱俊
ZHU Jun(Anhui Technical College of Water Resources and Hydroelectric Power,Hefei 231603,China)
出处
《安徽水利水电职业技术学院学报》
2023年第4期31-34,68,共5页
Journal of Anhui Technical College of Water Resources and Hydroelectric Power
基金
安徽省高校自然科学研究重点项目(2022AH052304,KJ2020A1041)。
关键词
损伤检测
小波包分解
梅林倒谱系数
主成分分析
神经网络
damage detection
wawelet packet decomposition
Merlin cepstrum coefficient
principal component analysis
neural network