Dealing with motion artifacts in long-term ECG recordings is a big issue. The frequency spectrum of motion artifacts is similar to the frequencies of the QRS complex--the wanted signal in the ECG. The deletion of motion artifacts often leads to a deformation of QRS complexes, too. These risks can be minimized by using a noise-correlating signal as a second channel for artifact reduction. This paper presents an approach using the electrode-skin impedance as a second channel for the reduction of motion artifacts. Using the discrete wavelet transform, motion artifacts can be deleted time and frequency selective. This filter approach leads to an improvement of the automatic QRS detection and decreases the number of false detections by 35 %.