Sensing is a fundamental process to acquire information in the physical world for computation. Existing models treat a sensing process as an indivisible whole, such that sampling and reconstructing of signals are designed to be highly associated with each other in a unified procedure. These strongly coupled sensing systems are efficient, but usually lack reusability and upgradeability. We propose a functional sensing model called SDR(Sampling-Design-Reconstruction) to decouple a sensing process into two modules: sampling protocol and reconstruction algorithm. The core of this decoupling is a design space, which is a common data structure constructed using functions of the sensing target as prior knowledge,to seamlessly bridge the sampling protocol and reconstruction algorithm together. We demonstrate that existing types of household electricity usage sensing systems can be successfully decoupled by introducing corresponding design spaces. Sensing is a fundamental process to acquire information in the physical world for computation. These models coupled a sensing process as an indivisible whole, such that sampling and reconstructing of signals are designed to be highly associated with each other in a unified procedure. These strongly coupled sensing systems are efficient, but are often lack reusability and upgradeability. We propose a functional sensing model called SDR (Sampling-Design-Reconstruction) to decouple a sensing process into two modules: sampling protocol and reconstruction algorithm. The core of this decoupling is a design space, which is a common data structure constructed using functions of the sensing target as prior knowledge, to seamlessly the sampling protocol and reconstruction algorithm together. We demonstrate that existing types of household electricity sensing systems can be successfully decoupled by introducing corresponding design spaces .