The 21-point average method

wavelet transform method and wavelet-packet transform method were applied to carry out de-noise of DTG curves for a kind of corn stalks sample

thus to obtain a more ideal DTG curve.Through comparing several parameters

like the peak characteristics

signal-noise-ratio(SNR)

signal-noise-ratio gain(GSNR)

time shifting

power spectrum

variation coefficient

the Lipschitz exponent

flatness index and others

the results show that

the wavelet-packet transform method prefers to keep the peak characteristics of the original DTG curve

and the de-noised DTG curve is smooth and has better de-noise effect.The noise filtered by the above three methods all have zero mean characteristics and independent distribution

which is similar to the Gaussian distribution.Considering the evaluation system with peak characteristics and smooth index

it helps to avoid the "less" or "over" smoothing phenomenon

and a more ideal de-noise curve can be obtained.