Abstract: A process and a device for automatically detecting an abnormal respiratory ound permits a continuous and stress-free registration of abnormal respiratory sounds of a test person under normal life activities over a period of time. The respiratory sounds are detected, processed and evaluated in real time. From the detected frequency range, in the course of the evaluation consideration is given only to the previously determined characteristic frequencies at the position of which in the frequency spectrum of the respiratory sound of the test person an abnormal respiratory sound is to be recognized by a significant feature. The data detected at these frequency positions are compared with data which are individually determined for the test person for fulfillment of specific criteria for the occurrence of an abnormal respiratory sound, and the result of the comparison is made available as a signal.