在未知訊號分離(BSS)領域中，許多研究者已經建立了很多不同的理論，而其中一種最廣為人知的就是獨立成份分析(Independent Component Analysis, ICA)方法。基於獨立成份分析(ICA)中的Kullback-Leibler divergence理論，可推導出一個近似函數，將此函數設為最佳化演算法中的評估函數，藉由最佳化演算法來將Kullback-Leibler divergence最小化，將可得到分離矩陣，透過分離矩陣可將混合在一起的訊號予以分離。 本研究應用粒子群優(Particle Swarm Optimization, PSO)演算法於未知訊號分離(Blind Signal Separation, BSS)問題中。比較了粒子群優(PSO)演算法與基因(Genetic Algorithm)演算法在未知訊號分離(BSS)中的效能。考慮三個線性混合的訊號當輸入，送進此系統中做計算，可成功分離出三個輸出訊號。比較輸出訊號與原始訊號，波形圖與平均運算時間可以顯示出此系統將訊號分離的效能。模擬結果顯示出粒子群優(PSO)演算法是一個強而有力的演算法，它的有效性以及計算效率很適合工程上的應用。

This paper applies Particles Swarm Optimization (PSO) algorithm to Blind Signal Separation (BSS) problems. In Blind Signal Separation field, many researchers have developed various kinds of theories. One of the most well-known theories is Independent Component Analysis (ICA). Based on the Kullback-Leibler divergence theory in Independent Component Analysis, one can derive an approximation function which is regarded as the fitness function in Optimization algorithm. By minimizing the Kullback-Leibler divergence via Optimization algorithm, we can have the demixing matrix which can separate mixed signals. We compares the performance of PSO algorithm and Genetic Algorithm in BSS problems. Considering three linearly mixed signals as inputs sent to the system for separation, three output signals can be separated. Comparing the output signals with source signals, the wave forms and average computational time illustrate the efficiency of signal separation of the system. The simulation results demonstrate that Particle Swarm Optimization is a powerful algorithm; its effectiveness and computational efficiency are suitable for engineering applications.