Enso play an role in the regulating interannual variability in tropics. However to elucidate other signal besides Enso, we need to filter out the signal. Is there any other known method other than using linear regression method?
There is an well-known alternative to regression for your question about signals, but its applicability for your requirements may depend a lot on the length and time-resolution of your data. By this I mean "spectral analysis" and possibly its later developments concerning wavelets. You will find the subject dealt with in mathematical/statistical books on "time-series analysis".You may find something on the meteorological literature dealing with cycles in long-record climate variables, although that type of research may have gone out of fashion recently. Since you write of "filtering out" a signal, there are indeed methods of doing this in signal processing that derive from spectrum analysis.
Parker et al. 2007, Zhang et al. 1999 frequency bandpass filter to filter out the signal. They used the period between approximately 2 and 6 years to identify the ENSO variability. However this assume that all of the SST variability in the chosen band, and none outside it, is associated with ENSO. This can be misleading as ENSO is known to have broader dynamics.
To filter ENSO out you first need to define what you mean by ENSO as you well pointed out. Yes, ENSO has a broad spectrum but you will have to make an assumption as to what scales to include in ENSO to be able to separate it. An alternative to linear regression or spectral analysis is using principal components. In our past work we used a "cannonical ENSO" as the leading complex principal component (EOF) in the 0–8-yr band of sea surface temperature anomalies (Mestas and Enfield, 2001, Eastern equatorial Pacific SST Variability: ENSO and Non-ENSO components..., J. Climate, 14, 391-402).