Timbre, as an element of musical expression, plays a vital role in performance, and is one of the elements that every performer can’t do without when playing. This paper establishes the string vibration model of guzheng music according to the principle of guzheng sound generation, and uses the fast Fourier transform to analyse the spectrum of guzheng audio. Then the discrete Fourier is used to replace the continuous Fourier change, and the signal is periodically diffused. Mel frequency cepstrum coefficients, which represent the characteristic parameters of guzheng timbre, are used to preprocess, fast Fourier transform, spectral line energy calculation, and Mel filter processing of guzheng music signals in order to obtain the discrete cosine transform. The harmonic structure is used as an objective index to define the guzheng timbre, for which the spectral coefficient of timbre expression is calculated to evaluate the quality of extracted timbre. The sound quality optimisation experiment is designed to analyse the time-frequency curve of guzheng and evaluate the sound quality optimisation effect. The amplitude of the time-frequency waveform signal at sampling point 2 of the guzheng is the largest, ranging from -0.75 to 0.75, and the duration of sampling point 3 is the longest, lasting nearly 4 × 104s. After optimisation, the performance of the timbre is the most significant in the high-frequency region of 1700 Hz, and the amplitude difference between the two sweeps of the curves fluctuates within the interval of [0.01,0.058], which confirms that the optimisation of the guzheng timbre quality is very good.