Generation of ATM Video Traffic Using Neural Networks E. Casilari, A. Reyes, A. Díaz-Estrella, F. Sandoval Dpto. Tecnología Electrónica E.T.S.I. Telecomunicación,Universidad de Málaga, Campus de Teatinos, 29071 Málaga, Spain casilari@dte.uma.es arcadio@dte.uma.es Abstract A new model to generate Asynchronous Transfer Mode (ATM) video traffic is presented. The model, implemented on neural networks, is capable of accurately adjusting the autocorrelation and probability distribution functions of a given video traffic. This adjustment is performed by capturing the projected conditioned histogram of the real traffic, so that the neural model will be able to yield a simulated as a function of an input white noise. Using neural networks we benefit from their inherent capacities for working in real time, because of their parallelism, and interpolating unknown functions. Results are presented for a real MPEG video source. 1 Introduction The aim of the future broadband integrated service digital network (B-ISDN), based on asynchronous transfer mode (ATM), is supporting a wide variety of multimedia services, with different statistical characteristics and quality of service requirements (delay, losses). Among the most emerging services, we have the video and image transmissions, which include many types of visual media, as still image, video-conferencing, broadcast TV, HDTV, etc.