Foetal heart rate estimation by empirical mode decomposition and MUSIC spectrum

It is still a challenge to estimate the foetal heart rate accurately from a strong nonstationary abdominal ECG signal. Even if signal process eliminates the predominant maternal ECG component, the foetal heartbeats are still very weak due to other existing interferences. This paper introduces empirical mode decomposition (EMD) and multiple signal classification (MUSIC) to tackle this issue. Firstly, preprocessing eliminates the interferences and noise in abdominal ECG signal and then the EMD is utilized to decompose the foetal ECG signal into a set of intrinsic mode functions, which could be used to detect the foetal QRS waves. Finally, the MUSIC is applied on the foetal QRS waves indicator sequence to estimate the foetal heart rate in the frequency domain with a high resolution. The basis functions of EMD are derived from the foetal signal under test, which makes the detection process robust and adaptive. In addition, the foetal heart rate estimation is carried out in the frequency domain regardless of the detection of R-wave peaks. In the simulated experiments with the proposed method, the mean value of the fHR estimation error is 2 BPM with a standard deviation of 1.5 at SNR = −30 dB, and it decreases to 0 when SNR = −16 dB. When compared to the fastICA algorithm, the proposed method shows robustness using three different real foetal ECG databases with variable degrees of nonstationarity.