Power consumption of signal-processing microcircuit systems has become increasingly important in recent years, primarily due to the explosion in demand for portable systems where battery-lifetime is a major marketing point. In addition, the thermal dissipation levels of microcircuits are beginning to affect reliability and future development. These factors have elevated the importance of power in the design process. The consideration of power adds another degree of complexity to the design process, hence the requeriment for power-conscious tools to support low-power system design. Considering power as a high-level parameter promises to maximise power reductions. However, the complex nature of the high-level design process is significantly increased with the consideration of power consumption; hence the interest in the application of heuristic optimisation techniques. This thesis presents a high-level tool for the power-conscious design of digital signal processing systems in CMOS technology. The tool consist of a specially tailored genetic algorithm with embedded high-level design properties. The embedding of these techniques required extensive modification to standard genetic algorithm operations to instil the algorithm with the ability to explore the low-power solution space efficiently. Problem specific genetic mutation and crossover operations were developed to incorporated high-level design transformations. To guide the search, power estimation strategies were examined and implemented within the fitness evaluation framework. The prototype system was extented to incorporate additional genetic techniques to improve the efficiency and results of the exploration. The power-estimation module was also enhanced to obtain estimations based on practical circuit synthesis techniques. The performance of the enhanced tool is illustrated with benchmark circuits, illustrating the tool's ability to reduce the power of practical signal processing examples. The multi-objective properties of the genetic algorithm are exploited to present design information illustrating the trade-off between area and power while keeping throughput constant. Alternative search techniques were developed and compared with the genetic algorithm. The results illustrate the superiority of the genetic algorithmin obtaining the lowest power solution and presenting trade-off information.