Driving behavior recognition is a notable topic in travel safety, as transportation and insurance companies could adopt effective tools to detect unsafe driving and internalize the associated costs. Different driving events and the related severity must be detected to distinguish abnormal behaviors. The global positioning system (GPS) provides useful information regarding the location of the vehicle at any time and is vastly used in various devices such as smartphones and GPS trackers. Other sensors, on the other hand, provide complementary valuable information but their implementation requires extra costs and more complex and intensive algorithms. We developed a threshold-based algorithm to detect the turning and braking of vehicles using the GPS sensor. The data contained 11 trips with a frequency of 1 Hz with a total duration of 2.7 h. The algorithm utilizes a supplementary map matching and a relabeling technique to boost the accuracy and yet preserve the reasonable computation load. The overall precision and recall rate of the turn-detecting model are respectively 77.5% and 92.5%. Also, this algorithm can detect braking events with a precision of 68.18% and a recall of 83.33%. To address the concerns about the overfitting, we tested our algorithm on a secondary dataset, and nearly similar values of accuracy were resulted, showing the flexible nature of our algorithm while dealing with a different set of driving behaviors and road characteristics. Additionally, a sensitivity analysis showed the sensitive nature of the brake detection algorithm, in contrast with the turn detection algorithm. Overall, our algorithm showed promising results and can be a pioneer one in the field of low-cost detection algorithms built for smartphones or GPS trackers possessed by various trucking and car insurance companies.