Software sensors (or observers) are especially useful when the state vector of a system cannot be measured. Widely employed for bioprocesses, they estimate the state variables from available measurements, which are functions of one or several state variables, provided that the observability property is checked. In food defrosting, an important challenge is to know the spread of the melting front inside the product, without any invasive measurement. The solution proposed in this paper is based on the design of software sensors able to estimate, from a superficial temperature measurement, the temperature distribution in the product, in real time, and consequently making it possible to determine the melting zone. The proposed solutions are, first, an extended Kalman Filter and, second, switched reduced Luenberger observers. The results are illustrated during the experimental thawing of a block of tylose inforced convection. (c) 2006 American Institute of Chemical Engineers.