The purpose of target detection is to detect and locate a set kind of target object from still pictures or videos. Most of the existing studies have simplified the target detection problem into a binary classification problem. Considering the superiority of support vector machine in pattern recognition, how to apply it to target detection has become the focus of computer vision. The traditional support vector machine is based on two kinds of problems, and how to effectively extend it to multiple kinds is still a problem to be studied. There are many unique advantages in solving small and nonlinear high-dimensional model recognition problems, and these problems can also be applied to other automatic learning problems, such as the attribute algorithm currently used to define models, bec feedback. The processing process of traditional virtual bass enhancement algorithm is analyzed, and it is found that the processing effect of different frequency components in the sound source is not uniform, which has a great influence on the final bass enhancement effect. Therefore, a new virtual bass enhancement algorithm is proposed based on the idea of sound source separation, which can reduce the distortion and improve the bass enhancement effect by separating the transient and steady components of the sound source before bass enhancement.