The Chirp signal parameter estimation has the problems of high computational complexity and low precision under low Signal-to-Noise Ratio (SNR). In order to solve these issues, an improved Chirp parameter estimation algorithm is proposed. According to the Chirp signal energy and the signal period, the square of the signal spectrum amplitude is inversely proportional to the frequency modulation slope. The Chirp signal is multiplied by the unknown Chirp signal to find the known Chirp signal with the largest square of the spectral amplitude. In order to achieve the parameter estimation with high accuracy, the Modified Rife (M-Rife) is used to estimate the center frequency and correct the amplitude of the multiplied signal. The algorithm is combined with the Short-Time Fourier Transform (STFT). First, STFT is used to estimate the approximate range and signal period of Chirp. The dichotomy is used instead of the equal step size to allow the improved Chirp signal parameter estimation algorithm to search within this approximate range, resulting in significant complexity reduction. The experimental results show that when the SNR is not less than -10 dB, the Normalized Mean Square Error (NMSE) of modulation frequency estimated by our proposed algorithm is better than Fractional Fourier Transform (FRFT) by 1~2 dB. When the signal-to-noise ratio is not less than -10 dB, the NMSE of center frequency estimated by our proposed algorithm is significantly better than FRFT and close to the Cramer-Rao Low Bound (CRLB) line.