Presently, the world of wireless communication is going under some crucial challenges, which attracts the attention of several researchers. Cognitive radio is defined as a multidimensional aware, autonomous radio system that learns from its experiences to reason, plan & decide future actions to meet user needs. Such a highly varied radio environment calls on intelligent management, allocation & usage of scarce resources. Issues like spectrum sensing & allocation, environmental learning i.e., adaptability & capability to learn attracts the attention of several soft computing learning & optimization techniques like neural networks, fuzzy logic, genetic algorithm & swarm intelligence. The cognitive engine behind the radio combines the sensing, learning, switching and optimization algorithms to control & adapt the radio system from the physical layer to the top of the communication stack. This paper presents a critical review on different soft computing approaches applied over the cognitive radio issues & also points out different research directions over it.