In the expansion of wearable devices, attitude estimation remains as one of the most important feature in the evergrowing selection of wearable devices, therefore, highlighting the importance of acquiring necessary knowledge on improving attitude estimation accuracy. Through the analysis and comparison of the cosine measure, quaternion algorithm, adaptive wave peak detection and the wearable human posture detection system in hardware selection and practical actualization, it summarizes the advantages and disadvantages of the four attitude judgment algorithms and their main applicable situations. Finally, a relatively optimized attitude detection method suitable for children bracelet is obtained.