Abstract:

In order to obtain the desired signal of dualaxis inclinometer from the complex automotive environment，a LMS（Least Mean Square） algorithm for reducing the convergence speed to process the related data is presented,it presents a self-adaptation step LMS algorithm by establishing the non-linear relationship between the step and the error,so as to the step is interrelated with the true signal for reducing the algorithm sensitivity to noise.To cope with the problem of superiority and correctness for applying to the vehicle vibration signal extraction,the learning curves of algorithm are gained by inputting the data with additive white noise of time domain.A comparison of them is made by convergence speed and computation complexity index.The result shows that the novel algorithm is reliable．The vehicle acceleration signal extraction experiment shows that it  removes the vehicle environmental impact of the dual-axis inclinometer,and  restrains its own output noise.
Based on the self-adaptive step LMS algorithm,it proposes and implements a scalable FIR （Finite Impulse Response）filter structure on FPGA(Field Programmabl
e Gata Array) .By reusing of fist-order filter unit to complement the multi-stage cumulative,it achieves 256 filter experiments on the largest FPGA chip.The result indicates that the FIR filter structure full use of chip resources,processing speed highly,reliability highly and apply for the low speed for vehicles attitude measurement system.

Key words: vehicle vibration signal, heavy background noise, signal extraction, least mean square (LMS) algorithm, variable step size, field programmable gata array (FPGA)implementation