Data Driven Approach for mmWave Channel Characteristics Prediction Using Deep Neural Network
| Autor: | M. Neema, E. S. Gopi |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Artificial neural network
Computer science business.industry Deep learning Convolutional neural network Autoencoder Computer Science Applications Data-driven Non-line-of-sight propagation Computer engineering Path (graph theory) Artificial intelligence Electrical and Electronic Engineering business Communication channel |
| Zdroj: | Wireless Personal Communications. 120:2161-2177 |
| ISSN: | 1572-834X 0929-6212 |
| DOI: | 10.1007/s11277-021-08768-7 |
| Popis: | The elaborate architectures used in millimeter wave (mmWave) MIMO communication system make its channel characteristics prediction difficult. In this regard, we consider the problem of predicting the path power loss for both line of sight and non line of sight mmWave channels by employing deep learning (DL) based data driven model. The success of a DL technique depends on how well we train the model, which again relies on the training data. It is often hard to get the amount of data needed to build and train a model. The paper proposes a methodology to predict the channel characteristics using the available limited data. For this, the work effectively develops two deep learning models: Model 1: Autoencoder for creation of new data from the limited data and Model 2: Deep Convolutional Neural Network (DCNN) for channel characteristic prediction. Simulation results show that the proposed model 1 is able to generate more data in the desired direction and model 2 can accurately determine the path power loss associated with a mmWave channel, with very minimum error. The paper also demonstrates a method of extracting useful information using the constructed data driven model. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full text from SpringerLink