نوع مقاله : مقاله پژوهشی
عنوان مقاله English
نویسندگان English
Audio signal processing in military environments has consistently faced significant challenges due to the presence of complex noises such as explosions, weapon fire, and electromagnetic interference. This study presents an advanced method for removing environmental noise from military audio signals, utilizing a Least Mean Squares (LMS) adaptive filter with a dynamic learning rate. The primary innovation of this research is the simultaneous use of five input signal features—namely, energy, variance, covariance, entropy, and zero-crossing rate—for the automatic adjustment of the learning rate. The optimal combination of these features was achieved using a genetic algorithm with a population of 50 chromosomes over 10 generations, yielding the optimal coefficient set: [0.8486, 0.1872, 0.5016, 0.9195, 0.3582]. Experimental results demonstrate that the proposed method achieves a 12.43 dB improvement in the Signal-to-Noise Ratio (SNR) – from -0.18 dB to 12.25 dB – and reduces the Mean Squared Error (MSE) to 0.0016. Furthermore, the method shows a 1.95 dB superior SNR improvement compared to fixed learning-rate filters and a 1.56-fold reduction in error variance. By maintaining an optimal balance between accuracy and computational complexity, this algorithm is well-suited for critical military applications.
کلیدواژهها English
