كنترل فعال كيفيت صدا توسط الگوريتم ضريب بهره پويا FxLMS با استفاده از تحليل زمان-فركانس براي صداهاي ناايستان

Today, sound quality control has become one of the fundamental challenges in various industries, including automotive, aerospace, an‎d home appliances, as the auditory experience of users directly influences their satisfaction an‎d perception of product quality. In this regard, this thesis develops a two-stage framework for active sound quality control of vehicle noise. In the first stage, an FFT DGF-FxLMS algorithm is designed for shaping stationary noise at constant intermediate speeds. To this end, sensitivity analysis is initially employed to identify key frequencies influencing psychoacoustic metrics. Subsequently, a multi-objective optimization process is used to determine the optimal amplitude an‎d relative phase values at these frequencies, defining a target signal with desirable acoustic parameters (loudness, roughness, sharpness, an‎d tonality). These optimal values are calculated for several reference speeds an‎d then generalized to intermediate speeds via an interpolation system. The results confirm the simultaneous improvement of all sound quality indices, where combined phase an‎d amplitude control plays a key role in enhancing sharpness. In the second stage an‎d as the main contribution, to overcome the limitations of the above method in non-stationary noise conditions, the STFT DGF-FxLMS algorithm is proposed for shaping non-stationary noise during run-up. This algorithm integrates a dynamic gain factor with Short-Time Fourier Transform (STFT) processing, enabling the tracking an‎d control of signals with time-varying characteristics. Simulation results demonstrate the success of both stages. Improvement of sound quality metrics at constant an‎d varying speeds, stable convergence of the advanced algorithm under non-stationary conditions whereas the FFT-based method diverges, an‎d significantly lower energy consumption of the proposed algorithm, requiring only 2% of the energy of a conventional ANE FxLMS system to achieve similar acoustic performance, are among the advantages of the proposed algorithm. Examination of spectrograms further confirms the concentration of energy at target harmonics an‎d the suppression of broadban‎d noise components. This research presents an efficient an‎d low-power solution for sound quality control in dynamic environments.

علي لقماني , مهدي كشميري

محمد دانش , سعيد ضيائي راد , روح اله طالبي توتي