In this paper, a hearing aid process is described, based on a composite filtering compressor, and its implementation, which is based on a digital hearing aid system. In the compressor, a relatively smooth compression curve similar to the FFT method is calculated using outputs of a conventional filter bank. Consequently, it is expected that the qualitative degradations in compressed speech are reduced and, as a result of computational reductions, the compression is suitable for DSP operations. Moreover, it is possible to realize an adaptive compression where the flame lengths for spectral estimation may be varied according to speech features. In order to implement the compression, a digital hearing aid system has been designed with some external adjustments, and the syllable articulation tests and the quality assessments have been carried out. As a result, it was confirmed that the appropriate compressions could be realized by the subjects' adjustments of the parameters.