1. Deep Ridge Regression Neural Network-based hybrid precoder and combiner design.
- Author
-
Nagapuri, Lalitha, Penchala, Suresh, Vallem, Sharmila, Navitha, C. H., and Rao, D. Sreenivasa
- Subjects
KRONECKER products ,CHANNEL estimation ,MIMO systems ,ENCYCLOPEDIAS & dictionaries ,ANTENNA design - Abstract
In mm-wave MIMO systems, hybrid precoder and combiner designs enhance antenna gain for improved transmission efficiency. However, beam-squint conditions during transmission impact throughput, affecting codebook and increasing beam focus and angle of arrival difference, degrading channel performance. Hence, a novel the Pylon ∂ PSO Method has been proposed to minimize codebook size and array gain, reducing the difference between beam focus and angle of arrival. A Grassmannian codebook is created without compromising throughput. For channel state estimation, existing techniques using the Kronecker product which face convergence errors due to improper hyperparameter matrix selection. Hence, an innovative Lagrange Dual technique and Separable K-Singular Value DE polymerization (K-SVDEp) have been used in dictionary learning that results in the Pt3 product to find the best dictionaries in which block sparse values are estimated using a Deep Ridge Regression Neural Network-based estimator that gives an optimum hyperparameter matrix and eliminates convergence error. Furthermore, designing a combiner from the hyperparameter matrix faces mathematical challenges. Hence, a novel Glasgow technique is utilized which converges the design parameter value with a local optimum obtained using the GEO algorithm. The proposed design has been implemented on the MATLAB platform and outperforms existing techniques with a high spectral efficiency of 45 bits/Hz, SNR of 13.4 dB, and low SER of 1 0 - 4 . [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF