量大

In Inverse Synthetic Aperture Radar ( ISAR) instantaneous imaging, to solve the problems of big computation burden and limited approach precision of adaptive chirplet transform, a fast algorithm is proposed.

在逆合成孔径雷达（ISAR）瞬时成像中，针对自适应chirplet分解方法运算量大，对信号的逼近精度有限等问题，提出了一种瞬时成像快速算法。

Traditional SVM trains Support Vector Machines classifier directly under linear condition to sort face images and non-face images, and for such kind of classifier, the training is difficult. The amount of compute is big and the speed of training is slow.

传统的SVM直接在线性条件下训练SVM分类器完成人脸与非人脸的分类，分类器训练困难，计算量大且速度慢。

In the audio and video decode modules, to improve the decoding efficiency, this thesis uses MMX and SSE instructions to realize the core algorithms.

在音频、视频解码模块中，针对解码算法运算量大，耗时长的特点，采用MMX和SSE指令实现核心算法，缩短了解码时间。