Filter bank multicarrier (FBMC) has emerged as the favourite multicarrier communication technique since it meets the demands of the next-generation standards, especially for communication under doubly dispersive channels. In order to combat the undesirable channel effects like inter symbol interference and inter carrier interference, the prototype filters used in FBMC systems must satisfy orthogonality conditions and be well localized simultaneously in both time and frequency domains. Even though conventional doubly localized filters are optimum in the absence of channel, they fail to satisfy orthogonality and localization criteria under doubly dispersive channels. In the present study, an adaptive filter design method is proposed in which the prototype filter is designed as the weighted sum of a time-limited Nyquist filter, a band-limited Nyquist filter, and doubly localized Gaussian filter based on the analysis of discrete ambiguity function (DAF). The filter is designed to vary its characteristics according to the maximum time and frequency domain dispersions of the channel, so that signal-to-interference ratio is maximized. In order to ensure distortion-less transmission, the perfect reconstruction criterion for distortion-less transmission is mapped to specific conditions for DAF according to two-dimensional channel scattering function. The superiority in performance of the proposed adaptive filter is analysed in terms of signal-to-interference ratio and bit error rate, and is compared with the conventional filters for various channel scattering conditions. © 2018, © 2018 IETE.