In order to suppress the influence of noise introduced in infrared hum idity detection based on tunable diode laser absorption spectrum (TDLAS) technique,multi-threshold wavelet denoi sing is firstly introduced on the differential signal between detection channel and reference channel,then the se cond-order harmonic component is extracted from the differential signal through the digital phase lock-in amplifier,and finally the humidity value is determined according to the amplitude of the second-order harmonic com ponent.The filtering procedure of multi-threshold wavelet denoising is proposed,and related experimental verif ication is carried out on the algorithm.The analysis results indicate that this technique can effectively sup press noises,and the second-order harmonic component can still be well extracted under extremely low humidity.By using the developed detection system and the prepared hermetic environment,the humidity measurement is perform ed thoroughly.Experimental results show that,by using multi-threshold wavelet denoising,the detection lim it can be dropped from 0.2% to 0.08%,the detection error under the humidity of 0.2% can be decreased from 48% to 7.9%,and that under other humidity is less than 5%.Because the noise is well suppressed through wavelet d enoising,the standard deviation of the system is minimized by one order of magnitude,and the detection stability is therefore improved.The wavelet denoising technique can also be adopted in other infrared detection systems with similar p rinciple for enhancing performance.