Wireless Sensor Network (WSN) is a centreless distributed wireless network constituted by mobile nodes without the help of any fixed infrastructure. Cooperation between nodes makes communication in WSN be achieved in complex environment with the deficiency of limited transmission distance made up by multi-hop forwarding so as to broaden the network coverage. The hop by hop pattern, however, may lead the connections between adjacent nodes lack of guarantee. The negative impact of communication performance in WSN caused by the limited bandwidth, high BER, fading, noise, interference, is alleviated by a novel adaptive filtering strategy based on frequency subbands and predetermined threshold, which is being obtained in Gaussian and the multipath fading channel according to frequency-matching principle and BER performance. The dynamic selection of subbands will obtain high use efficiency without the help of frequency hopping, and propound a new thought to improve band limited communication for WSN. The effectiveness of the adaptive filtering method has been verified by ISS-OFDM in different interference conditions, and the simulating results based on NS2 indicated that the system BER can be improved by 5dB at most.