In this paper, a model-based adaptive filter is used to suppress electrical noise in a high-frequency noninvasive valvometry device, which is part of an autonomous biosensor system using bivalve mollusks valve-activity measurements for ecological monitoring purposes. The proposed model-based adaptive filter uses the dynamic regressor extension and mixing method to allow a decoupled estimation of the parameters. Once the desired regression form of the output model is obtained, a fixed-time estimation approach is used to identify its parameters. By applying these two techniques, a flexible filter structure is obtained with the property of retaining the major relevant components of interest of the original valve-activity signals, even in the case when the unwanted signal frequency components are in the same frequency range as the useful variables.