This paper addresses the problem of speaker emotion recognition. Modeling various emotions is challenging due to their complex and non-linear characteristics. The Polynomial Chirplet Transform (PCT) is particularly well-suited for capturing such emotional variations in speech signals. This paper designs a new algorithm for estimating the optimal polynomial functions for PCT when analyzing single signals. Since a dataset generates numerous polynomial functions, some of which may be noisy, the paper employs a density-based clustering algorithm to identify the optimal set of polynomial functions. Using these functions, a feature extraction system called Multi-PCT is constructed. Experimental results demonstrate that the proposed features outperform previous methods across several evaluation metrics.