**3. Conclusions**

The characteristics of frontopolar-temporal EEG signals of depression patients are investigated using signal processing techniques and nonlinear parameters. EEG signals for the analysis were acquired from 30 unipolar depression patients and 30 age and sex matched healthy controls. Bipolar EEG recording using a 24-channel EEG machine was carried out at locations FP1-T3 (left half) and FP2-T4 (right half) of the brain for a duration of 5 minutes each, under eyes closed and eyes open condition in a resting state. Total variation filtering was found to be effective in removing the high frequency noise while the eye blink and eye movement artefacts were removed by visual inspection. Wavelet analysis is performed and signal features having significant influence on the signal waveforms of depression and normal controls have been identified. Coiflet 5 is used for the wavelet analysis. An 8-level decomposition was carried out and Relative Wavelet Energy was calculated on the reconstructed signal coefficients. Wavelet Entropy calculations revealed the degree of disorder associated with the EEG signals. The nonlinear measures like Approximate Entropy (ApEn), Fractal Dimension (FD) and Largest Lyapunov Exponent (LLE) are calculated. Nonlinearity of the EEG signal under study was confirmed by surrogate data analysis.

Depression effects are reflected mainly in the lower frequency range indicating a reduced brain activity. The multiresolution decomposition characterised the various frequency bands of EEG signals. The wavelet energy distribution in different frequency bands indicated higher levels of brain activity for normal controls in gamma, beta and alpha bands. It also showed lower brain activity in the delta bands of depression patients. The results from the calculations of RWE confirm the fact that, mental activity as reflected in the EEG signals of depression patients is confined to the lower frequency range especially in the delta band of 0-4 Hz. The quantitative evaluation of nonlinear parameters like ApEn, FD and LLE confirmed higher brain activity for normal controls compared to depression patients. Lower values of these nonlinear parameters indicate the fact that complexity of EEG is reduced in depression which effectively helped in discriminating EEG signals of depression patients and healthy controls. The quantitative assessment of signal characteristics and nonlinear parameters from the present study may be of significant use in the analysis of brain dynamics.
