Channel Impulse Response Peak Clustering Using Neural Networks

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216305%3A26220%2F23%3APU151103" target="_blank" >RIV/00216305:26220/23:PU151103 - isvavai.cz</a>

Výsledek na webu

<a href="https://commnet-conf.org/6thEditionProceedings/index.html" target="_blank" >https://commnet-conf.org/6thEditionProceedings/index.html</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1109/CommNet60167.2023.10365265" target="_blank" >10.1109/CommNet60167.2023.10365265</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Channel Impulse Response Peak Clustering Using Neural Networks

Popis výsledku v původním jazyce

This paper introduces an approach to process channel sounder data acquired from Channel Impulse Response (CIR) of 60GHz and 80GHz channel sounder systems, through the integration of Long Short-Term Memory (LSTM) Neural Network (NN) and Fully Connected Neural Network (FCNN). Theprimary goal is to enhance and automate cluster detection within peaks from noised CIR data. The study initially compares the performance of LSTM NN and FCNN across different input sequence lengths. Notably, LSTM surpasses FCNN due to its incorporation of memory cells, which prove beneficial for handling longer series. Additionally, the paper investigates the robustness of LSTM NN through various architectural configurations. The findings suggest that robust neural networks tend to closely mimic the input function, whereas smaller neural networks are better at generalizing trends in time series data, which is desirable for anomaly detection, where function peaks are regarded as anomalies. Finally, the selected LSTM NN is compared with traditional signal filters, including Butterworth, Savitzky-Golay, Bessel/Thomson, and median filters. Visual observations indicate that the most effective methods for peak detection within channel impulse response data are either the LSTM NN or median filter, as they yield similar results.

Název v anglickém jazyce

Channel Impulse Response Peak Clustering Using Neural Networks

Popis výsledku anglicky

This paper introduces an approach to process channel sounder data acquired from Channel Impulse Response (CIR) of 60GHz and 80GHz channel sounder systems, through the integration of Long Short-Term Memory (LSTM) Neural Network (NN) and Fully Connected Neural Network (FCNN). Theprimary goal is to enhance and automate cluster detection within peaks from noised CIR data. The study initially compares the performance of LSTM NN and FCNN across different input sequence lengths. Notably, LSTM surpasses FCNN due to its incorporation of memory cells, which prove beneficial for handling longer series. Additionally, the paper investigates the robustness of LSTM NN through various architectural configurations. The findings suggest that robust neural networks tend to closely mimic the input function, whereas smaller neural networks are better at generalizing trends in time series data, which is desirable for anomaly detection, where function peaks are regarded as anomalies. Finally, the selected LSTM NN is compared with traditional signal filters, including Butterworth, Savitzky-Golay, Bessel/Thomson, and median filters. Visual observations indicate that the most effective methods for peak detection within channel impulse response data are either the LSTM NN or median filter, as they yield similar results.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

20202 - Communication engineering and systems

Projekt

<a href="/cs/project/GF23-04304L" target="_blank" >GF23-04304L: Vícepásmová predikce šíření milimetrových vln pro dynamické a statické scénáře v členitých časově proměnných prostředích</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2023

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

2023 6th International Conference on Advanced Communication Technologies and Networking (CommNet)

ISBN

979-8-3503-2939-1

ISSN

—

e-ISSN

—

Počet stran výsledku

7

Strana od-do

1-7

Název nakladatele

Institute of Electrical and Electronics Engineers Inc.

Místo vydání

Maroko

Místo konání akce

Rabat, Maroko

Datum konání akce

11. 12. 2023

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—