This paper examines using a genetic algorithm to optimize singular value decomposition (SVD) for signal detection in cognitive radio networks. SVD is used to detect the presence of wireless signals by factorizing the received signal matrix. A genetic algorithm is used to optimize the number of columns in the covariance matrix input to SVD. Simulation results show the genetic algorithm optimized SVD method provides better detection probability than SVD alone, especially in low signal-to-noise ratio environments. The genetic algorithm optimized SVD is suitable for blind spectrum sensing where the properties of the signal to be detected are unknown.