Impedance plethysmography (IP) has long been used to monitor respiration, although the applied use of this technology has been limited to hospitals and other medical facilities. The largest obstacle that arises when one uses the IP signal for portable monitoring of patients is susceptibility to motion artifact. A small movement of the arm may result in an artifact whose amplitude is several times larger than the amplitude of the respiratory component of the signal. Furthermore, the frequency spectrum of the artifact often includes low frequencies, overlapping with that of respiration. As a result, the conventional motion artifact reduction methods such as adaptive filtering and independent component analysis prove to be inadequate means of extracting the respiratory component of the signal. This paper proposes a novel motion artifact reduction method that adopts Vapnik’s idea of epsilon-tube to restrict the amplitude of the filtered signal. Additionally, it uses Stockwell Transform to maximize the regularity of the filtered signal within the tube. The results indicate that epsilon-tube filtering has significantly higher accuracy compared with the currently existing motion artifact reduction methods. The combination of a portable IP device and the proposed epsilon-tube filter will allow the physicians to more accurately and adequately monitor patient respiratory activity within the home setting and without the use of facemasks or nasal cannulas, thereby not imposing any restrictions on patient’s airways.