Automatic modulation format bit rate classification and signal quality estimation in wireless communication systems using pattern recognition techniques/Teow Chiew Hoon
Terdapat pelbagai jenis modulasi dan kadar bit dalam persekitaran tanpa wayar yang berlainan ciri-ciri perambatan. Penerima boleh menerima data secara tepat tanpa sebarang pengetahuan dari isyarat penghantar dengan menggunakan proses yang bernama pengelasan modulasi automatik (AMC). Pengenalanpastian jenis modulasi dan kadar bit boleh membantu penerima dalam demodulasi isyarat dengan tepat. Penyelidikan projek ini adalah mengenai pengelasan format modulasi/kadar bit automatik dan anggaran kualiti isyarat dalam sistem komunikasi tanpa wayar dengan menggunakan teknik pengesanan corak. Teknik analisis komponen prinsipal (PCA) adalah dicadangkan untuk mengekstrak sesetengah ciri-ciri penting data yang digunakan oleh pengelas berasaskan rangkaian neural buatan (ANN) secara berkesan bagi pengelasan format modulasi/kadar bit automatik dan anggaran nisbah isyarat-hingar (SNR). PCA boleh mengurangkan dimensi data dengan berkesan tanpa kehilangan maklumat yang banyak. Pengelas berasaskan ANN latihan dan ujian serta pemilihan ANN parameter juga dibincangkan dalam projek ini. Satu data yang besar mengandungi plot tidak segerak (ADTPs) dari enam isyarat yang berbeza kadar bit/format modulasi dan SNR yang berlainan telah dihasilkan. Ciri-ciri penting ADTPs yang dihasilkan dengan menggunakan PCA dapat mengurangkan dimensi data sebanyak 90%, menyebabkan pemprosesan data dengan lebih mudah. Ciri-ciri penting data ini digunakan dalam latihan kedua-dua pengelas berasaskan ANN. Salah satu ANN adalah untuk pengelasan format modulasi/kadar bit automatik dan satu lagi ANN adalah untuk anggaran SNR. Prestasi ANN yang baik juga dihasilkan dengan berjaya melalui mengoptimumkan ANN parameter. Selepas latihan ANN tamat, pengelas ini boleh digunakan untuk pengelasan format modulasi/kadar bit automatik dan anggaran SNR bagi isyarat yang tidak diketahui dari penghantar. Pengelas berasaskan ANN mempunyai ketepatan yang tinggi, iaitu 98.18% bagi pengelasan modulasi automatik dan ralat purata SNR adalah 1.09 dB. Hal ini terbukti bahawa pengelas berasaskan ANN dalam projek ini bukan sahaja membolehkan pengelasan kadar bit dan format modulasi automatik secara tepat, malah ia juga boleh menganggar kualiti isyarat dari segi SNR.
___________________________________________________________________________________
There are varieties of modulation schemes and data rates in a wireless environment with diverse propagation characteristics. A receiver can receive data correctly without any prior knowledge of transmitted signal by using a process called Automatic Modulation Classification (AMC). The identification of the modulation type and data rate can help the receiver to demodulate the signal accurately. This research project focuses on the investigation of automatic modulation format/bit rate classification and signal quality estimation in wireless communication systems by using pattern recognition techniques. The research considers the use of Principal Component Analysis (PCA) for automatic extraction of important signal features which can be effectively exploited by the developed artificial neural network (ANN)-based classifiers for joint bit rate and modulation format classification as well as signal-to-noise ratio (SNR) estimation. PCA can reduce the data dimensions effectively without much loss of information. The training and testing of ANN-based classifiers and the selection of optimum ANNs’ parameters are also presented in this project. A large data set containing asynchronous delay-tap plots (ADTPs) for six different commonly used modulation formats and bit rates of the signals is generated for several different SNR values. The critical features of these ADTPs are then obtained by applying PCA which also reduces the dimensionality of data by approximately 90%, thereby making the subsequent processing of data much easier. The extracted features are utilized for the training of two ANN-based classifiers. One ANN is trained for the classification of modulation formats and bit rates while the other ANN is trained for the estimation of SNR of the signals. During the training process, various parameters of ANN-based classifiers are optimized for good performances. Once the training is complete, these classifiers are used for the automatic classification of modulation formats and bit rates of signals as well as for the estimation of SNR of unknown signals without any prior information from transmitters. The developed ANN-based classifiers demonstrate high accuracies i.e. an accuracy of 98.18% for automatic modulation format and bit rate classification and an average SNR error of 1.09 dB for SNR estimation. This proves that ANN-based classifiers developed in this project can not only enable accurate automatic bit rate and modulation format classification but can also provide information about the quality of signal in terms of SNR.
Automatic modulation format bit rate classification and signal quality estimation in wireless communication systems using pattern recognition techniques/Teow Chiew Hoon
Automatic modulation format bit rate classification and signal quality estimation in wireless communication systems using pattern recognition techniques_Teow Chiew Hoon_E3_2014_