This paper discusses the design and prototyping of a cognitive radio system architecture that enables for TVWS exploitation under the real time secondary spectrum market policy. It describes a centralized infrastructure-based network, where TVWS allocation among unlicensed systems is administrated by a spectrum broker, carrying out radio-resource management and spectrum trading in real time. For optimum system performance as a matter of maximum-possible radio resource exploitation and trading revenue, the paper studies and implements a prototype mechanism at the spectrum broker side, which exploits backtracking algorithm for obtaining the best-matching solution. Performance evaluation experiments carried-out under controlled conditions verified the validity of the proposed architecture, besides establishing its capacity for maximum spectrum utilisation and minimum fragmentation under a fixed-price trading policy.