It is evident that the 3G(3rd-generation) and B3G(Beyond Third Generation in mobile communication system) communication systems need higher wireless capacity. Multi-user detection(MUD) technique is one of the key techniques in Code Division Multiple-Access(CDMA) system. This technique can reduce the multiple access interference (MAI) and inter symbol interference (ISI) by using the structured information at the receiving end in the CDMA system, and it can also realize the information demodulation while there exist multi-user signals. This paper presents a new Multi-user detection technique, that is simulated annealing-genetic algorithm. This proposed algorithm applies simulated annealing (SA) to genetic algorithm (GA). First, the initial group is given, then the genetic algorithm is used to make the search gradually convergent to the feasible globally optimal solution or sub-optimal solution, then the simulated annealing is used to do the local optimization. Simulations show that the proposed multi-user detector is close to the optimal one and better than the conventional one in the performance of anti-MAI and anti-ISI.