Deep neural network a posteriori probability detector for magnetic recording

Unmet Need: Multi-track detection for two-dimensional magnetic recording

Hard disk drives (HDDs) shipped currently use one-dimensional (1D) perpendicular magnetic recording. HDD industry is approaching an areal density limit for conventional 1D magnetic recording. HDDs store only binary data, hence equalization techniques such as OFDM are not suited. HDD recording channels have media noise. Currently trellis based Viterbi algorithm (VA) is used to detect coded data bits in presence of ISI and media noise. Two-dimensional magnetic recording suffer complexity with VA and it grows exponentially with channel length and predictor order.

The Technology: Using deep neural network a posteriori probability for two-dimensional magnetic recording

A novel deep neural network based a posteriori probability detection system with parallel multi-track detection for two-dimensional magnetic recording channels on next generation hard disk drives is proposed by WSU researchers.


  • Two-dimensional magnetic recording of hard disk drives.


  • Detects coded bits in presence of intersymbol interference, intertrack interference and media noise.
  • Support wide variety of applications through general network structures and training techniques.
  • System achieves 3% and higher areal density gain over VA for various data sets.

Learn More

Scott Steiger
Associate Director
Washington State University
(509) 335-7065
Reference No: TECH-19/3280


Krishnamoorthy Sivakumar
Benjamin Belzer
Jinlu Shen
Ahmed Medhat Aboutaleb

Key Words

Machine Learning
Two Dimensional Magnetic Recording