A patient with hearing loss mainly in the high-frequency range can not clearly comprehend speech. Therefore, if the frequency band is compressed to a lower range, it is possible that the speech perception of such a patient can be improved. To test this possibility, the instrument described below was constructed and experimentally tested. First of all, a speech sample was divided into seven frequency bands, and the zero-crossing waves and envelopes were determined. The zero-crossing waves were converted to lower frequencies by flip-flop circuits, and they were amplitude-modulated by their respective envelopes. Frequency-compressed speech was obtained by the addition of the seven signals thus processed. Successful training of a person with normal hearing ability was conducted using this frequency-compressed speech. It was concluded that a high intelligibility can be thus obtained. We expect to employ this instrument in the treatment of patients with hearing impairments.