A physiological sensor with personalized biological feedback is disclosed, wherein the electrocardiogram (ECG) signal-capturing module detects the electrocardiogram (ECG) in real-time, and detects the tested persons motion state by a tri-axial accelerator automatically. At the same time, the voice-playing module can vary the proper music by the personalized learning process according to the different users different motion states, so that the music style can fit the motion type at that time. In addition, when the intensity of exercise exceeds the intensity that can be afforded by the heart, the earphone voice prompt can immediately give the voice suggestion to delay the violent exercise temporarily, thereby preventing the potential danger of over-violent exercise.