The invention relates to a multi-
hypothesis code rate adaptive system and method oriented to an HTTP-based dynamic adaptive streaming medium. The 
system comprises a 
server, a channel terminal and a 
client which are connected in sequence, the 
client comprises an MPD (
Multimedia Presentation Description) file acquisition analysis module, an 
adaptive method set, a controller, a 
code rate set module and a buffer region, the 
adaptive method set is connected with the controller, and the controller, the 
code rate set module and the buffer region are connected in sequence. Each 
adaptive method operates a code rate request 
algorithm, and the adaptive algorithms operated by the adaptive methods are different. According to the 
system and the method, the long-time 
quality of experience of a user reaches the optimal; multiple single rate adaptive methods can be integrated, the methods are complementary in advantages and mutually independent, and the universality and the 
operability are achieved; and compared with the conventional code rate 
priority method and the method based on a proportional plus derivative controller, the return value of the method is higher, and the defect of extremely instable return in a start stage of the method based on 
reinforcement learning is overcome.