A Frequency-Time Algorithm of Parameter Estimation for Sinusoidal Signal in Noise

Abstract

In this paper, a computationally efficient and high precision parameter estimation algorithm with frequency-time combination is proposed to improve the estimation performance for sinusoidal signal in noise, which takes the advantages of frequency- and time-domain algorithms. The noise influence is suppressed through spectrum analysis to get coarse frequency, and the fine frequency is obtained by denoising filtering and using linear prediction property. Then, estimation values of the amplitude and initial phase are obtained. The numerical results indicate that the proposed algorithm makes up for the shortcomings of frequency- and time-domain algorithms and improves the anti-interference performance and parameter estimation accuracy for sinusoidal signal.

A Parameter Estimation Algorithm for Damped Real-value Sinusoid in Noise

“…where the hat above the parameter stands for the estimated value, and Re[ ] and Im[ ] denote a real and a complex part of a complex number, respectively. Furthermore, we can obtain the amplitude and phase according to (4).…”

“…Parameter estimation of damped real-value sinusoidal signal in noise is a basic but significant problem in signal processing. It is used in many areas, such as signal spectrum analysis, power quality detection systems, instrument measurement devices, and others [1]- [4]. For instance, the free damped vibration signal of the flow tube of a digital Coriolis Mass Flowmeter (CMF) can be used to track the natural frequency change of the flow tube, and the natural frequency is used to drive the vibration of the flow tube [5]- [7].…”

A high‐precision frequency estimation algorithm based on multi‐segment signal fusion

“…Recently, based on the excellent properties of autocorrelation signal, Chen et al. [20, 21] proposed a fusion algorithm, which performs autocorrelation and averaging on multi‐segment signals, and then using expanded autocorrelation (EA) or modified covariance (MC) algorithm [22, 23] for frequency estimation. Compared with other algorithms in time domain, the accuracy is improved, and the method is easy to implement, but the disadvantage is that the accuracy drops significantly at low SNR.…”

“…Then Chen et al [19] carried out coherent averaging on multisegment signals in the time domain to improve the SNR and used FFT for frequency estimation, but the method needs to ensure that each segment of the signal has the same initial phase, which is difficult to implement in practice. Recently, based on the excellent properties of autocorrelation signal, Chen et al [20,21] proposed a fusion algorithm, which performs autocorrelation and averaging on multi-segment signals, and then using expanded autocorrelation (EA) or modified covariance (MC) algorithm [22,23] for frequency estimation. Compared with other algorithms in time domain, the accuracy is improved, and the method is easy to implement, but the disadvantage is that the accuracy drops significantly at low SNR.…”