Blind adaptive energy estimation for decorrelating decision-feedback CDMA multiuser detection using learning-type stochastic approximations

This paper investigates the application of linear reinforcement learning st ochastic approximation to the blind adaptive energy estimation for a decorr elating decision-feedback (DDF) multiuser detector over synchronous code-di vision multiple-access (CDMA) radio channels in the presence of multiple-ac cess interference (MAI) and additive Gaussian noise. The decision feedback incorporated into the structure of a linear decorrelating detector is able to significantly improve the weaker users' performance by canceling the MAI from the stronger users. However, the DDF receiver requires the knowledge of the received energies, In this paper, a new novel blind estimation mecha nism is proposed to estimate all the users' energies using a stochastic app roximation algorithm without training data. In order to increase the conver gence speed of the energy estimation, a linear reinforcement learning techn ique is conducted to accelerate the stochastic approximation algorithms. Re sults show that our blind adaptation mechanism is able to accurately estima te an the users' energies even if the users of the DDF detector are not ran ked properly. After performing the blind energy estimation and then reorder ing the users in a nonincreasing order, numerical simulations show that the DDP detector for the weakest user performs closely to the maximum likeliho od detector, whose complexity grows exponentially with the number of users.