In this paper, we study the scale-transformation weak signal detection method based on chaotic Duffing oscillator. Based on this, the frequency characteristics of the weak single-frequency signal with arbitrary frequency and initial phase can be extracted. Furthermore, we propose a signal frequency interception preprocessing method for analyzing weak multi-frequency signal. After combing the preprocessing method with the weak signal detection method based on chaotic Duffing oscillator, we further propose a novel detection method for weak multi-frequency signal. Based on this novel method, we can extract the frequency characteristics and initial phase characteristics of the weak multi-frequency signal. In this research, we also study the automatic detection of unknown multi-frequency signal. According to the numerical simulation, the novel method we propose in this paper is helpful to extract the frequency parameters and initial phase information of each signal component of the weak multi-frequency signal.