 This paper presents a novel approach to calibration-free recognition of multi-frequency modulated visual stimuli. The approach uses a phase difference constraint to control the phase difference between the spatially filtered stimuli, which is then used as input to a canonical correlation analysis, CCA, algorithm. This allows for improved recognition accuracy compared to existing methods, while also reducing the computational complexity. This article was authored by Qi-Man Wang, Zeyu Wang, Boyu Wang, and others.