In electro-optical tracking system,fractal noises and white noises have main effects on the precision of inertial sensors,and it is difficult to filter fractal noises by traditional methods.Because of special properties of fractal noises in wavelet transform domain,denoising means with wavelet transform are adopted and data fusion of sensors is implemented.By simplifying the thresholds computation and mending the symmetric extension for signals input,the processing precision is improved and calculation length for signal is reduced greatly.Therefore,the signal can be processed in real time.Some experiments for inertial sensors are carried out and the experimental results verify that the algorithm is valid and real-time.