Oscillation has become one of the important problems faced by modern power grids. Multi-types of oscillations may occur simultaneously in the power system and the oscillation frequency span is extremely large. For signals with wide-band oscillation modes, the signals in different frequency bands are first separated by a band-pass filter, and then the Improved Variational Mode Decomposition (IVMD) method with high noise robustness is used to extract each oscillating mode signal. Finally, the combinations of Hankel total least squares (HTLS) and adaptive neural network algorithm (Adaline ANN) is used to estimate the frequency, attenuation factor, amplitude and phase of low-frequency oscillations. Furthermore, the introduction of Adaline neural network solves the problem that the mode amplitude and phase are difficult to determine after IVMD processing, so that the detection accuracy is improved. Simulation and case analysis show that this method can effectively distinguish and extract different types of oscillation modes in the signal, and accurately identify the information of each mode. The IVMD-HTLS-Adaline method can effectively identify signals that have experienced severe oscillations or noise-like signals with potential oscillations.