针对传统的小波去噪方法容易产生信号振荡和 丢失特征信息等问题，提出了一种基于蝙蝠算法的新的有效的红外光谱去噪方法。该方法创新 性地运用蝙蝠算法优化了小波阈值和估计因子。其基本思想是，首先在解空间中随机生成一定规模 的个体，然后根据向当前最优个体学习的方法进行速度更新，从而实现位置更新；同时，由于Lévy 飞行搜索策略会产生较大跳跃，利用这种不均匀、随机游走的特性可以实现对整个解空间的搜索，从而避免陷入 局部极值点。CO气体红外光谱去噪实验的结果表明，利用蝙蝠算法对各个分解层的阈值和估计因子 进行优化后，信噪比为84.184，均方误差为0.0006。由于更有针对性地保留了光谱信号中的特征 信息并剔除了无用的噪声信息，该方法可以提高后续定性和定量分析的精度。

To solve the problems present in traditional wavelet denoising methods, such as unexpected oscillation occurring and loss of characteristic information, a novel and effective wavelet denoising method for IR spectrum based on bat algorithm (BA) is proposed. In the method, both threshold and estimation factor are optimized by BA innovatively. Its basic idea is that firstly a certain size of individuals are generated in the solution space randomly and then the velocity and location of each bat are updated according to its distance from the best bat individual. Meanwhile, by using the random walking characteristics of Levy flight search strategy, the search for the whole solution space can be implemented and falling into the local minimum can be avoided. The experimental result of CO gas IR spectrum denoising shows that after the threshold and estimating factor in each layer of wavelet decomposition are optimized by using the proposed wavelet denoising method, the signal-to-noise ratio (SNR) is up to 84.184 and the root square error (RMSE) is 0.0006. Because the characteristic information in the spectral signal is reserved and the unwanted noise information is removed more accurately, the method can be used to improve the accuracy of subsequent qualitative and quantitative analysis.