A signal processing method and a hearing device for isolating a user's own voice from microphone signals is disclosed. The signal processing method comprises the following steps. Providing at least two microphones at one ear of a user or at least one microphone at each ear of the user. Receiving sound signals by the microphones. Isolating a user's own voice from the microphone signals by way of a combination of at least two of the following processes: a) applying own voice detectors to the microphone signals, b) using monaural and binaural directional microphone systems, c) using monaural and binaural directional microphone systems with near field emphasis, and d) applying speaker identification models to the microphone signals. Estimating a stress level or cognitive load from the isolated voice. Adjusting settings of a device in dependence of the estimated stress level or cognitive load.
