In this paper we study, through theoretical analysis and simulation, the impact of Beacon Order (BO) and Superframe Order (SO) parameters of IEEE 802.15.4 on the networks performance and we investigate their optimal values for different classes of traffic. The traffic is dimensioned according to the requirements of the CANet project in which a cane becomes a mean of communication and a surveillance system embedding several sensors to monitor the elderly health and environment(voice, pressure, temperature, etc.). The cane's sensors impose different QoS constraints. Depending on the expected throughput, a sensor's traffic will fall within one of three classes that we defined. Therefore, in order to ease the understanding of our optimization, we introduce a classification scheme which applies to the existing quality of service algorithms. We derive by theoretical study the optimal values of BO and SO that should be used to fit each traffic class QoS requirements and we validate our results by simulation.