| Heart rate,as one of the four physiological indexes of the human body,changes in its value can reflect the health degree of the human heart in daily life and exercise.The use of wearable devices can help people monitor heart rate anytime and anywhere without being restricted by the use scene.At present,most wearable devices on the market calculate heart rate values based on photoplethysmography(PPG).The heart rate is calculated by measuring the attenuation degree of light passing through human blood,tissues and other structures to reflect the heart rate.However,the PPG signal itself is extremely susceptible to interference from various noises.In addition to system noise,it also contains noise such as motion artifacts.Under normal circumstances,the greater the degree of exercise and the greater the amplitude of human body vibration,the easier it is to produce a relative displacement between the wearable device and the skin.The motion artifacts generated at this time will mask the effective heart rate information contained in the PPG signal.Comparing the calculation accuracy of the heart rate of different types of exercises,it is found that the noise interference generated by strenuous running is the strongest,the artifacts in the PPG signal are difficult to remove,and the calculated heart rate often differs from the real heart rate too much.Therefore,it is an urgent problem to extract effective heartbeat information and improve the accuracy of heart rate calculation in a strong noise interference environment.In response to the above problems,this paper proposes a heart rate calculation scheme based on PPG signals in a strong noise interference environment.The processing flow is mainly divided into two parts: PPG signal processing based on zero-phase filtering and Wiener filtering,and adaptive generalization.The main tasks of peak tracking model heart rate extraction include:1.In the PPG signal processing stage: This paper proposes a method for removing motion artifacts based on zero-phase filtering and Wiener filtering.First,eliminate the interference of phase offset on the preprocessed PPG signal through zero-phase filtering,and solve the problem of heart rate spectrum peak offset caused by band-pass filtering,and then use the three-axis acceleration as the noise reference in the filtering process of the Wiener filter.To remove the motion artifacts mixed in the PPG signal,and get the clean PPG signal as much as possible.2.Spectrum peak tracking stage in the frequency domain: A heart rate extraction algorithm based on an adaptive spectrum peak tracking model is proposed.Through the combined use of cadence information and historical heart rate information,the heart rate peak in the current time window is dynamically determined.The possible appearance range of the heart rate peaks is limited,and the accuracy of the heart rate peaks identification is improved.In solving the problem of determining the model parameters,the improved particle swarm optimization is used to perform offline learning on the test data,and the model parameters with strong applicability are obtained,which improves the effect of the model.In the experimental verification stage,in order to better reflect the accuracy of the heart rate algorithm proposed in this article,the 2015 IEEE signal processing cup public data and the selftest data evaluation algorithm collected by the experimental equipment in this article are used.The experimental results show that compared with the current common heart rate algorithms,the heart rate estimation method proposed in this paper can obtain a calculated heart rate with less error according to the prediction range given by the adaptive model when the heart rate peak is completely covered by the noise peak.To avoid excessive fluctuations in the heart rate calculation.According to the application results of actual products,it can greatly improve the accuracy of heart rate calculation in the application of smart armband devices. |
PDF Full Text Request