Abstract : This PhD thesis presents new solutions for power management in autonomous microsystems. The autonomous microsystems are the new candidates for the nodes of wireless sensor network. These systems contain microgenerators that harvest ambient energy and convert it into electricity. This thesis has been done in the framework of a European project (VIBES : VIBration Energy Scavenging FP6 IST-1-STREP-507911). This project focused on harvesting energy from mechanical vibrations. In this case piezoelectric or electromagnetic microgenerators are used for the conversion. The microgenerators are connected to a storage element by the way of power management modules. The main challenge for the design of these modules is the low voltage (few millivolts) and the ultra low power (hundreds of nanowatts) provided by the microgenerators. To overcome this problematic three approaches are proposed. The first technique allows the increase of the output voltage given by the microgenerator. This technique is called SSH (Synchronized Switch Harvesting). The efficiency of this technique was validated with a piezoelectric macrogenerator. The influence of scale down effects on this technique is studied, and a more suitable solution is proposed for microgenerators. The second approach is based on a voltage multiplier. This circuit plays the role of AC/DC and DC/DC converters. It accepts input voltages in the order of tenth of millivolt. This passive circuit uses diodes with very low threshold voltage. The third approach deals with the design of ultra low power AC/DC converter (tenth of nanowatt). This converter can rectify signals of few millivolts. The three approaches are implemented with two types of integrated-circuits technologies.