Optimal Design of Tunable Bandpass Audio Filter using Variable Learning Rate Backpropagation Algorithm

Artificial intelligence techniques are increasingly being employed for solving optimization problems due to their simplicity and speed. And artificial neural networks and their many variants are unsurprisingly at the forefront of this growing trend. In this paper, a variable learning rate backpropagation algorithm (VLRBA) was deployed to train a designed tunable bandpass audio filter for optimal performance. Different configurations of the neural network were evaluated and the best performing network was selected for the task. The performance of the neural network was analyzed using standard metrics such as regression and mean square error. The high training regression value of 0.99328 and testing regression value of 0.98381, achieved after 129 iterations, demonstrate the effectiveness of the variable learning rate backpropagation model in optimizing the tunable bandpass filter circuit. This, therefore, represents an important addition in the quest for faster and efficient circuit designs, especially in the sound engineering domain.