Energy efficient equalizer design for MIMO OFDM communication systems using improved split complex extreme learning machine

• Published in: Signal, image and video processing Vol. 16; no. 2; pp. 349 - 357
• Main Authors: Sahoo, Swetaleena, Sahoo, Harish Kumar, Nanda, Sarita
• Format: Journal Article
• Language: English
• Published: London Springer London 01.03.2022; Springer Nature B.V
• Subjects: Adaptive algorithms; Artificial neural networks; Communications systems; Computer Imaging; Computer Science; Equalizers; Image Processing and Computer Vision; Machine learning; MIMO communication; Multimedia Information Systems; Original Paper; Orthogonal Frequency Division Multiplexing; Pattern Recognition and Graphics; Quadrature amplitude modulation; Signal to noise ratio; Signal,Image and Speech Processing; System effectiveness; Vision; Doppler shift; ALM; MIMO OFDM; SCELM; Eye diagram
• ISSN: 1863-1703; 1863-1711
• DOI: 10.1007/s11760-021-01933-2

Design of effective equalizer for modern multiple-input multiple-output orthogonal frequency division multiplexing (MIMO OFDM) wireless system is an important research problem as reconstruction of original information is quite challenging in the presence of different channel impairments. The equalizer design addressed in this paper uses IEEE 802.11 g indoor channel model and quadrature amplitude modulation (QAM) constellation to describe the fading statistics with a limited user mobility. The important research contributions can be visualized looking into two important aspects which includes a split complex extreme learning machine model trained using adaptive Levenberg–Marquardt algorithm and ON/OFF strategy of the equalizer using threshold based on signal to noise ratio (SNR) and inter symbol interference that indirectly helps in designing an energy efficient wireless receiver. Evaluation parameters like mean square error (MSE), eye diagram and symbol error rate (SER) have been considered in this paper to prove the efficiency of the proposed algorithm. The MSE value decreases to the range of 10 - 2 with a varying training testing (I/K) ratio of 0.05 to 0.5. In addition, the proposed equalizer model outperforms the other existing methods in terms of SER by revealing a low value of 0.001 in 7 dB SNR in indoor channel conditions.