Abstract: To solve the problem that the existing loudness compensation algorithm in digital hearing AIDS cannot meet the speech intelligibility and audibility of hearing impaired patients with low high frequency sensitivity, a nonlinear frequency compression method of digital hearing aids based on the auditory characteristics of human ear and cochlea was proposed. First, according to the characteristic that the first two formants in speech play a key role in vowel recognition, and the formant detection of speech signal was carried out. The first two formants were kept in the uncompressed frequency band to determine the total compression ratio of frequency. Then, the frequency domain was converted to the Mel domain, and the ratio of each subband in the Mel domain to the total frequency band to be compressed was calculated by using the feature that the human ear in the Mel domain has an approximate linear perception of the Mel frequency. As the basis for the distribution of the compression ratio of each subband in the frequency domain, nonlinear frequency compression was performed in the frequency domain. Finally, multi-channel loudness compensation was performed for the low-and medium-frequency signals that are sensitive to hearing impairment after frequency shift compression according to the hearing curve of hearing impairment patients. Results show that compared with the comparison algorithm, the proposed algorithm can better protect the spectral peak ratio of speech, and improve the audibility and intelligibility of patients with hearing impairment.