| Vital signals such as respiration and heartbeat are of great significance in observing the state of human life.Non-contact life signal detection technology has extensive applications in emerging fields such as anti-terrorism detection,medical monitoring,disaster rescue,and autonomous driving.The life signal detection technology based on frequency modulated continuous wave(FMCW)radar offers advantages such as non-contact operation,large-scale detection capability,anti-interference ability,multi-parameter monitoring,and real-time performance.It holds significant potential in healthcare,safety,and intelligent transportation,providing enhanced protection for people’s lives and health.However,the signal processing process of FMCW radar is relatively complex,involving steps such as false alarm judgment,signal processing,and frequency extraction.Precise signal processing and algorithm design are required to obtain accurate life signal information.To address the challenges of background clutter interference and the separation of respiratory and heartbeat signals in FMCW radar life signal detection,this paper proposes a respiratory signal processing algorithm based on the GOCFAR Kalman cascade.The algorithm incorporates false alarm judgment,Kalman filtering for signal processing,and respiratory signal frequency extraction and calculation.Through simulation experiments and measured data experiments,the GOCFAR Kalman cascade algorithm proposed in this paper achieves a 52% reduction in average absolute error and a 4.9 percentage point decrease in the average absolute error ratio,compared to other traditional detection algorithms.In addition,in order to solve the problem of poor heartbeat signal detection performance of the cascade algorithm,this paper introduces a robust adaptive filtering algorithm based on the least mean square(LMS)method to extract heartbeat signals.The algorithm adopts a robust multi-stage LMS adaptive filtering model with fast convergence speed and small mean square error(MSE).It is composed of two modules.Module 1 uses the multistage adaptive filter structure to get the estimated value of life signal,module 2 uses the input signal to restore the pure signal.The results of simulation and experimental data show that the multilevel cascade robust adaptive filtering algorithm based on LMS can effectively improve the heartbeat signal detection performance.Compared with other traditional detection algorithms,the average absolute error is reduced by 58% and the average absolute error ratio is reduced by 6.1 percentage points.To improve the anti-noise performance and generalization ability of FMCW radar,this article finally introduces deep learning technology and proposes a deep learning life signal detection algorithm based on the Long Short-Term Memory Network(LSTM)for radar signal processing.The algorithm leverages a large amount of data to train the model,fine-tune network parameters, and verify model performance through a validation set to prevent overfitting.It utilizes the LSTM neural network to unwrap one-dimensional phase information from radar signals and extract frequency information to determine the frequency of life signals.Through testing ten groups of measured data,the algorithm demonstrates minimal detection error,with an average absolute error of 6.0358 for respiratory signals and 4.56 for heartbeat signals. |
