本論文乃是針對Ｍ準位頻移鍵（MFSK，M-ary frequency-shifted keying）調變信號在相加性白色高斯雜訊（AWGN，additive white gaussian noise）通道中之調變形態之識別進行研究。
　　整個研究過程分為兩個階段。第一階段為信號模擬及數據擬合（fit）階段。本階段的目的在透過電腦模擬，（即所謂的蒙地卡羅（Monte Carlo）法），隨機產生大量的MFSK信號樣本，輸入AWGN通道後，在接收機端萃取樣本之瞬時頻率，並統計瞬時頻率的機率分佈，利用最小平方法擬合分佈的數據，以尋求其瞬時頻率的機率密度函數。模擬結果顯示，MFSK信號之瞬時頻率機率密度分佈可由Tikhonov函數表示。
　　第二階段為性能分析及識別驗證階段。依第一階段求得的瞬時頻率機率密度函數，推導出測試統計並依據之發展出識別器的架構。其後並分析該識別器之性能。最後吾人以BFSK/4FSK為例，計算在不同SNR下之識別成功率。當SNR大於0dB以上時，識別成功率高於99.8%，顯示本識別器性能良好。

　　In this thesis we present a modulation classifier for classifying the M-ary frequency-shifted keying (MFSK) signal corrupted by the additive white Gaussian noise.
　　There are two stages during the development of the classification algorithm. At the first stage we extract the instantaneous frequency from raw data which is generated by the computer simulation program. Next the histogram is drawn by using the instantaneous frequency samples. It is found that the resulting histogram can be viewed or approximated to be the Tikhonov function in which the parameter must be modified. The omdified Tikhonov functions are used to be the probability densities for M-ary FSK signals and will be the key role for deriving the required test statistics in the second stage.
　　In the second stage we develop the test statistics, then employ the derived test statistics to construct an MFSK modulation classifier. We give an example to demonstrate the capability of BFSK/4FSK classifier in term of the probability of successful classification. The results via computer simulations show that successful rate of the proposed classifier is higher than 99.8% when SNR>10dB.