 This paper presents a new method for detecting peak error negativity, PEN, and EEG signals related to cognitive conflict processing. The method uses a combination of multi-scale local maxima extraction and rotation to identify PEN candidates, which are then further processed using shape and location-based features to eliminate false positives. The proposed method has been tested on EEG data collected during a 3D object selection task, showing improved accuracy over existing methods and requiring less than 4 milliseconds per epoch to process. This makes it suitable for use in BCI applications. This article was authored by Tranheepdin, Avinash Kumar Singh, Gwin Lintrung, and others.