Filter bank multi carrier (FBMC) is considered a competitive waveform candidate for 5G that can replace orthogonal frequency division multiplexing (OFDM). However, channel estimation (CE) is a big challenge in FBMC because it suffers from intrinsic interference which is due to the orthogonality of the subcarrier functions in the real field only. In this paper, we investigate a proposed modified interference approximation scheme (M-IAM) by approximating the intrinsic interference from the neighboring pilots to accommodate the complex channel frequency and thus improving CE performance besides simplifying its processing. The M-IAM scheme has larger pseudo pilot magnitude than other conventional preamble schemes, namely the interference approximation method (IAM) with its versions (IAM-C) and (E-IAM-C); in addition to the novel preamble design (NPS). In addition, the proposed (M-IAM) scheme is characterized by the lower transmitted power needed. The CE performance of the M-IAM is investigated through 512 and 2048 subcarriers via different types of outdoor and indoor multipath fading channels that are time-invariant such as IEEE 802.22, IEEE 802.11, Rician, and additive white Gaussian noise (AWGN), as well as time varying channels such as Rayleigh and Vehicular A (Veh-A). Simulation results demonstrate that the proposed M-IAM scheme achieves a lower bit error rate (BER), lower normalized mean square error (NMSE) and lower peak-to-average power ratio (PAPR) over the conventional preamble schemes under the aforementioned channel models. The proposed scheme has the advantage of saving the transmitted power, a requirement that could match 5G low power requirements.