Heart rate and oxygen saturation are two common vital signs that provide crucial information about a person's cardiovascular and respiratory health. Nowadays, mobile applications can measure heart rate as well as oxygen saturation by contact photoplethysmography (PPG) method. However, there is still room for improvement in existing mobile applications for vital sign measurement using contact photoplethysmography due to the issues of motion artifact and fixed thresholding. Thus, this paper proposes an algorithm for heart rate and oxygen saturation measurements based on contact PPG method and video streaming using noise reduction and adaptive thresholding approach for accuracy performance improvement. The noise reduction concept is applied to the heart rate measurement algorithm in terms of parameter adjustment to minimize the motion artifact. Whereas, for oxygen saturation measurement, an adaptive thresholding algorithm is implemented together with machine learning algorithms, including decision tree classification and linear regression. Furthermore, data collection process is conducted on cardiopulmonary patients with the purposes of device calibration as well as functionality verification and accuracy performance benchmarking with commercial pulse oximeters. Clinical testing proves that the overall average accuracy of the algorithm has achieved 93.73% for heart rate measurement and 96.79% for oxygen saturation measurement, respectively. Additionally, this heart rate and oxygen saturation measurement function is integrated a six-minute walk test monitoring system as a test application for home rehabilitation.