Adaptive LMS Algorithm for Effective Noise Suppression

Noise cancellation aims to reduce or eliminate unwanted signals from the original signals, such as audio signals, thereby improving signal quality and clarity. A popular method for active noise cancellation is the adaptive filter, which uses an algorithm to dynamically adjust the filter parameters based on the input signal to effectively remove noise. An adaptive filter, a type of digital filter, adjusts its coefficients to reach an optimal state without needing prior information. The Least Mean Squares (LMS) algorithm is a well-known adaptive filtering technique utilized in signal processing and communication systems. This algorithm is a form of stochastic gradient descent used to minimize the mean square error between the desired signal and the filter's output. To assess the performance of the proposed LMS-based noise cancellation algorithm, both subjective and objective analyses were conducted. The subjective evaluation involved Mean Opinion Score (MOS) tests to gauge the perceived quality of the speech signal under various Signal-to-Noise Ratio (SNR) conditions in noisy environments. Objective evaluation was performed using Segmental SNR. The results from the listening tests showed the following MOS scores for the processed speech "The burse can lead on a smooth plank": 3.3, 3.5, 4.3, 4.0, and 4.3 for SNR values of 0, 5, 10, 15, and 20 dB, respectively. In contrast, the MOS scores for unprocessed speech were significantly lower: 1.0, 1.5, 2.0, 2.0, and 2.8 for the same SNR values. Additionally, Segmental SNR values for the processed speech improved to 4.5712, 4.0248, 3.5847, 3.0045, and 2.547 for the corresponding SNR values. These findings indicate a noticeable improvement in speech quality under different SNR conditions due to the LMS algorithm. Furthermore, similar subjective and objective analyses were performed for various speech signals in the presence of car noise and babble noise, demonstrating the algorithm's effectiveness across different noise environments.