微型电脑应用

为了降低手环故障率，提高智能手环使用寿命，设计了一种神经网络模型用来预测智能手环的故障时间与交互方案。通过代表性与多样性序列（RDS）体系，选取最能体现手环运行特征的时序序列，再基于分段特征表示，从中选取包含故障信息的时序事件序列。然后将上述数据导入长短时记忆（LSTM）网络，通过高斯混合模型去除手环冗余数据，并采用变分模态分解（VMD）对剩余数据进行降噪，最后Bi-LSTM网络与LSTM网络融合实现手环的故障预测。经过实验证明，LSTM神经网络在多方面可实现手环的故障精准预测。

To reduce the failure rate of smart bracelets and extend their service life,a neural network model is designed for predicting the fault time of smart bracelets and developing an interaction scheme.Through the representative and diversifying sequences(RDS)evaluation system,time series that best represent the operating characteristics of the bracelets are selected;then,based on segmented feature representation,time series event sequences containing fault information are selected from these time series.Subsequently,the above data are input into the long short-term memory(LSTM)network.Redundant data of the bracelets are removed using the Gaussian mixture model(GMM),and the remaining data are denoised by the variational mode decomposition(VMD)algorithm.Finally,the Bi-directional LSTM(Bi-LSTM)network is fused with the LSTM network to realize fault prediction of the bracelets.An interaction interface for the bracelet based on fault prediction is designed.Experimental results show that the LSTM neural network can achieve accurate fault prediction of the bracelets in multiple aspects.