Motor current signature analysis (NICSA) provides a method for non-intrusive and continuous monitoring of motors in many applications. Initially, the fast Fourier transform (FFT) was the main signal processing method used in such analysis. Recently, wavelet packet decomposition (WPD) has become popular in such applications. The second approach gives better results in the case of non-stationary signals but it has much higher computational complexity limiting its use with real time power quality monitoring devices. Use of four-channel FIR cosine-modulated filter banks is proposed in this study to provide lower computational complexity.