An experimental study was carried out to develop a new knock-detection method using cylinder pressure, block vibration and sound pressure signals from a spark-ignition (SI) engine. As a first step for knock detection, the 1st, 2nd and 3rd harmonic knock frequencies were determined by analyzing the cylinder pressure signals under a wide range of engine operating conditions. Since these knock frequencies were found to be independent of engine operating conditions and the standard deviations of these knock frequencies were small, three narrow band-pass filters were designed and applied to process the raw signals from the engine. Then a new method, called the Sum of Divided Band-Pass (SDBP) filtering method, was proposed to determine knock intensity, and it was shown to be better than the other methods in terms of knock detection sensitivity and background noise removal.Knock intensity greatly varies with engine operating conditions, knock sensor locations and fuel characteristics, etc. In order for a knock control system to work properly in a vehicle, it is required to set up a proper threshold value that can be applied under any circumstances. This ‘universal’ knock threshold value was found by analyzing the non-dimensional knock intensity distribution obtained by a non-dimensional version of the SDBP method which is called as NSDBP method. Also it was confirmed that this non-dimensional knock threshold value is independent of background noise level and sensor mounting locations.The new knock-detection method proposed in this study, consisting of the NSDBP method and the non-dimensional knock threshold value, will contribute to improve future knock control systems of SI engines.