 An ultra-accurate gate phase estimator has been developed by training a neural network on data collected from the hip and knee joint angles of 14 participants while they walked on a treadmill or overground. The collected data included normal walking speeds between 0.1 meters per second and 1.9 meters per second, as well as various conditions such as long strides, short strides, asymmetrical walking, stops, and sudden speed changes. The spatial analysis showed that the RMSE was 1.74 degrees and 2.35 degrees for the treadmill and overground walking, respectively. The temporal analysis revealed that the D67 detected heel strike events with an average MAE of 1.70 degrees and 2.74 degrees for the treadmill and overground walking, respectively. The performance of the D67 was consistent across all participants and conditions. Further analyses indicated that the D67 was able to accurately estimate the gate phase even when there were variations in stride length, limp, and sudden starts or stops. Additionally, the D67's performance was compared to other techniques, which confirmed its superiority and compar. This article was authored by Muhammad Shastri, Hannah Dinovitzer, Jia Qingwang, and others. We are article.tv, links in the description below.