In order to suppress sensor noise with unknown statistical properties in an electrical domain, a novel mid-infrared CH4 sensor was proposed based on a 3 291 nm interband cascade laser and a multipass gas cell. This sensor operates using Recursive Least Square (RLS) self-adaptive denoising algorithm and Direct Laser Absorption Spectroscopy (DLAS) technique. Based on the traditional detector output (called the signal channel), a noise channel was added to generate electrical noise using the feedback signal of the laser driver. Numerical simulation in MATLAB was performed to evaluate the filtering performance of the RLS algorithm in the DLAS application. By adding different noise into the driving signal of the laser, potential denoising and sensing capabilities of the RLS algorithm were experimentally evaluated. Taking into account only the intrinsic noise of the sensor, the Allan deviation indicates a measurement precision of ~78.8 nL/L) with a ~6 s averaging time without using any filter. For comparison, Allan deviation of ~43.9 nL/L was obtained with a ~ 6 s averaging time using self-adaptive filtering. The reported sensor, incorporating the RLS self-adaptive denoising algorithm, demonstrates enhanced noise immunity and sensitivity compared to the mid-infrared DLAS sensor using the traditional sensing architecture and the filtering method.