Deep Learning Networks for Optimal Power Compensation in IR-UWB Channel

Abstract

Considering the characteristics of wireless signal amplitude distribution in Impulse Radio Ultra-wideband (IR-IR-UWB) fading channel, the Automatic Gain Control (AGC) loop of IR-UWB communication system can obtain the optimal reference power to maximize the signal-to-noise ratio of the output signal of analog-to-digital converter (ADC) in AGC loop. In a multipath channel, the channel impulse response of IR- UWB signal arrives in clusters, which is different from OFDM signal amplitude obeying Rayleigh distribution, the arrival time of pulses in each cluster obeys Poisson distribution, and the amplitude obeys exponential distribution. Because of characteristics of IR-UWB signal distribution, under the condition of certain signal power, the different AGC reference power will make different ADC sampling noise power. An AGC optimal reference power is obtained by analyzing ADC sampling noise power with different reference power when the ADC sampling output SNR is maximum. According to IEEE 802.15.3a channel model, 4 different conditions IR-UWB channel impulse response amplitude distributions are simulated in this paper. Using normalization method, the average ADC sampling output SNR with different AGC reference power is simulated, and the maximum SNR result corresponding to reference power is the optimal one. An AGC optimal reference power with different ADC parameters is obtained by analysis method based on amplitude distribution. The IEEE 802.15.3a channel model is classified by a deep Convolutional Neural Network (CNN) so as to obtain channel model parameter under different channel impulse response. The simulation results show that the CNN has a high classification probability for 4 different channel model and an AGC loop stably outputs at the optimal reference power.