Abstract: Starting with the significance of the stellar research, the basic conception and importance of stellar atmosphere parameters are introduced. The two major classes of estimation approaches of stellar atmosphere parameters, the direct technique and the indirect technique, are discussed. This paper focuses on the indirect technique, which can be further divided into several methods:the photometric method, the infrared flux method, the Balmer profile method, the spectral line ratio method, the line index method, the metal line diagnostics method, the spectral template fitting method and the machine learning method. While the spectral template fitting method is widely used in current large spectroscopic survey projects, such as SDSS, RAVE and LAMOST, the machine learning method has shown its advantages in many specific situations. Among various machine learning methods and technologies, the most popular ones are principal component analysis (PCA), k-nearest neighbors (KNN), support vector machine (SVM), and all kinds of artificial neural network (ANN) methods. On the other hand, traditional astrophysical methods still play prominent roles. The metal line diagnostics method is still favored by astronomers when analyzing high resolution spectra. The experimental results from infrared flux method are generally used as calibration.