Healthcare is one of the important applying fields of wireless sensor network. The technologies of body Sensor Network based healthcare system include wireless sensor network, wired network and long-distance wireless communication. And design of small size, low consumption, low price sensor node is prerequisite for the technique. The BSN healthcare system is characterized by monitoring paitent influence the patients' activities of daily living relative to other healthcare devices. This feature brings convenience to users, and also more challenge for sample of vital signals and automatic diagnosis algorithm. It also brings the problem of activity recognition. According to above description, a healthcare platform is designed based on body sensor network. ECG characteristic points recognition algorithm, activity recognition algorithm is designed. The work mainly consists of the following aspects: Firstly, an improved architecture of body sensor network is put forward. The body sensor network based on this architecture would decrease the bandwidth of single user. Secondly, a body sensor network system is designed based on MC13213 SOC processors, PXA270 Xscale processor and IEEE 802.15.4 protocol stack. The sensor node has characteristics as follows: low power consumption, small size and extensibility. Thirdly, IEEE 802.15.4 MAC protocol is deeply analyzed. The MAC parameter suitable to healthcare system is proposed based on simulation. Application layer protocol and program is designed. Fourthly ECG denoising algorithm and characteristic points algorithm is research based on multi-class wavelet. The algorithm of wavelet denoising uses a lifting scheme wavelet transform. So the algorithm has characteristics with fast calculating speed, less memory. So the denoising algorithm is favorable to running in embedded system. Fifthly, Activity of daily living recognition used one triaxial accelerator sensor is designed based on multi-class support vector machines algorithm.