The invention relates to an improved electrocardiosignal
noise reduction method based on
singular spectrum analysis. The method specifically comprises the following steps: selecting a proper window length for electrocardiosignals to construct a track matrix, and carrying out
singular value decomposition; calculating a
signal reconstruction component corresponding to each
singular value, and removing a trend term from the first reconstruction component by adopting a low-order polynomial fitting method; adopting a wave trap to remove
power frequency interference in each reconstruction component; calculating an absolute value of a
correlation coefficient between the reconstruction components and comparing the absolute value with a threshold value to select an effective
signal component; and finally, superposing the effective
signal components to obtain a denoised electrocardiosignal. The method is driven by data, does not need reference input, is suitable for removing
Gaussian white noise,
baseline drift, myoelectricity interference,
power line noise and other
noise, and can improve the subsequent signal analysis effect. According to the method, the signal
distortion degree is small,
noise reduction can be well achieved under the condition of the low signal-to-noise ratio, and the method can be popularized to
noise reduction of one-way or multi-way other physiological signals.