This paper deals with the design of the low complexity and efficient dynamic spectrum learning and access (DSLA) scheme for next-generation heterogeneous decentralized Cognitive Radio Networks (CRNs) such as Long Term Evolution-Advanced and 5G. Existing DSLA schemes for decentralized CRNs are focused predominantly on the decision making policies which perform the task of orthogonalization of secondary users to optimum vacant subbands of fixed bandwidth. The focus of this paper is the design of DSLA scheme for decentralized CRNs to support the tunable vacant bandwidth requirements of the secondary users while minimizing the computationally intensive subband switchings. We first propose a new low complexity VDF which is designed by modifying second order frequency transformation and subsequently combining it with the interpolation technique. It is referred to as Interpolation and Modified Frequency Transformation based VDF (IMFT-VDF) and it provides tunable bandpass responses anywhere over Nyquist band with complete control over the bandwidth as well as the center frequency. Second, we propose a tunable decision making policy, ρt_randρt_rand, consisting of learning and access unit, and is designed to take full advantage of exclusive frequency response control offered by IMFT-VDF. The simulation results verify the superiority of the proposed DSLA scheme over the existing DSLA schemes while complexity comparisons indicate total gate count savings from 11% to as high as 87% over various existing schemes. Also, lower number of subband switchings make the proposed scheme power-efficient and suitable for battery-operated cognitive radio terminals.