Research Of Airport Noise Identification And Its Solidification On FPGA

As a symbol of modern city, while the civil aircraft provides the developing city with convenient transportation, it also leaves behind serious noise pollution at the same time. Therefore, how to evaluate the effects of aircraft noise on the surrounding residents is particularly urgent. Recognizing the aircraft noise event accurately is very important to the civil aviation airport noise monitoring and its evaluation. It not only establishes a foundation for assessing the aircraft noise pollution, but also can avoid waste of resource caused by noise monitoring equipment.There is no mature aircraft noise identification at present, most noise identification equipments are based on PC, high power consumption, bulky volume, mobile hard, easy to crash and expensive. Field programmable gate array(FPGA) has the characteristics of small volume, low power consumption, fast computing speed, stable performance, low cost and repeatable programming, which is an ideal alternative to noise identification equipment. On the base of previous work, the 1/3 octave spectrum data extraction process is explored, and the practical effect of a noise events is displayed. Then root mean square method is improved and tested, the results turn out that the noise feature is bell-shaped distribution, and it illustrates the feasibility of compacting the noise feature at the same time. By comparing artificial neural network and support vector machine in implementation principle, advantages and limitations, actual application effect, the identification of realizing the least squares support vector machine in the form of artificial neural network is established. As the implementation of the least squares support vector machine involves nonlinear kernel functions, in order to adapt to the kernel function in larger range and high resolution, the implement of floating-point multiplication, addition and subtraction operation is introduced, and resource usage is analyzed too. After that, the design of solving the least squares support vector machine's parameters and its simulation is given. At last, the parameters obtained is written into FPGA directly on the basis of support vector machine classifier model designed by DSP builder, realizing the design of recognizing aircraft noise events in real-time.Simulation results suggest that noise recognition implementation on FPGA has a high real-time performance, which can successfully meet the need of real time recognition of aircraft noise. Combined with characteristics of FPGA, we can draw a conclusion that the least squares support vector machine realized on FPGA in the form of artificial neural network is a set of reliable and cheap aircraft noise identification scheme.