Two-dimensional data detection and error control

Data detection refers to the process by which a continuous-valued signal, appearing at the output of a storage channel is converted into a discrete-valued signal representing an estimate of the original (possibly encoded) stored data. Implicit in this definition is the notion that channel outputs are continuous valued. This will be true even in cases for which the storage medium itself may be highly nonlinear or even bistable. A magnetic storage medium, for example, may have only two stable directions of magnetization; however, the readout process is necessarily imperfect (e.g., fringing fields, low-pass electronics, noise) resulting in a continuous-valued readout signal [1]. This continuous-valued signal serves as the input to the data detection algorithm whose job is to use its prior knowledge of the various corrupting influences of the channel along with its knowledge of the alphabet from which the channel inputs were drawn, to generate an estimate of the stored data. It is important to note that the continuous-valued channel output contains everything we can know about the stored data. The process of converting continuous-valued retrieved signals into discrete-valued code words or user symbols is, therefore, critical in terms of insuring the highest possible data fidelity.