Research on Speech Enhancement Algorithm based on Objective Evaluation

Different speech enhancement algorithms perform differently in different noise environments, and have their own advantages and disadvantages. By reasonably evaluating the performance of speech enhancement algorithm in different noise environments, it is helpful to improve the algorithm and select suitable application scenarios. In this paper, speech enhancement algorithms based on different principles are considered. For white noise, factory noise and coffee shop noise, three representative speech enhancement algorithms are selected, including frequency domain wiener filtering and improved subspace method. PESQ objective evaluation method is used to evaluate the performance indicators of different speech enhancement algorithms in terms of signal to noise ratio improvement and objective evaluation.