正交分頻多工系統(Orthogonal Frequency Division Multiplexing, OFDM)中,載波頻率偏移(Carrier Frequency Offset, CFO)所導致之頻率不同步,會破壞子載波間之正交性而產生載波間干擾(Inter-carrier Interference, ICI),因而降低系統效能。另外,上行正交分頻多重存取系統(Orthogonal Frequency Division Multiple Access, OFDMA)由於基地台同時接收多個用戶之上傳訊號,因而須承受來自不同用戶的載波頻率偏移(CFOs),故會產生嚴重的子載波間干擾(ICI)與多用戶干擾(Multiple Access Interference, MAI),此將對CFOs的估計帶來更大的挑戰。使用最大概似法(Maximum Likelihood, ML)雖然可得到最佳的效能,但因複雜度過高而難以實現,為了在效能和複雜度之間取得平衡,故一般文獻都採用次佳化演算法來實行,例如簡化型最大概似法(Simplified Maximum Likelihood, SML)和快速演算法(Fast Algorithm),雖然這些方法效能不錯,但仍存在運算複雜度過高的問題。 為了降低運算複雜度,本研究應用粒子群演算法(Particle Swarm Optimization, PSO)來估計CFO。然而隨著用戶增加,PSO會因多樣性(Diversity)不足導致效果遞減。因此,在PSO中結合突變(Mutation)機制,增加族群個體多樣性,擴大搜尋範圍,並進一步加入田口運算(Taguchi Method)提升PSO的區域搜尋能力來改善效能。實驗結果顯示我們提出的方法之效能可以接近SML及Fast Algorithm,且在大部分的情況下可以大幅降低運算量。
For the orthogonal frequency division multiplexing (OFDM) system, carrier frequency offset (CFO) leads to loss of orthogonality among subcarriers; thus significantly degrading symbol error performance. Moreover, in the uplink orthogonal frequency division multiple access (OFDMA) system, multiple CFOs due to multi-user result in multiple access interference (MAI), further degrading system performance. Maximum Likelihood is recognized as the best solution, but its computational complexity is too high to be implemented for practical applications. To achieve a best compromise between system performance and complexity, many suboptimal algorithm, such as Simplified Maximum Likelihood and Fast Algorithm, have been presented. While those methods demonstrate well performance, their computational complexity are still very high. For reducing the computational complexity, this study employs PSO algorithms to estimate CFO, however, its performance gradually degrades with the increasing number of users due to the lack of diversity. Therefor, this study embeds mutation operation and Taguchi method in the PSO to compromise the capability of PSO in local search and global search. Experimental results indicate that the proposed approach can achieve the same performance as SML and Fast Algorithm with less computational complexity in most cases.