A real-time multi-channel EEG signal processor based on on-line recursive independent component analysis, wherein a whitening unit generates covariance matrix by covariance computing according to the received signals sampling, the covariance matrix can calculated by an inverse square root matrix SVD unit to produce whitening matrix, an ORICA calculation unit can get whitened sampling signal by calculating sampling signals and whitening matrix, and get the independent components information by independent component analysis that uses the whitened sampling signal with an unmixing matrix, an ORICA training unit trains the unmixing matrix based on the independent components information to generate a new unmixing matrix, and the new unmixing matrix will send to the ORICA calculation unit for the computation of independent component analysis next time. The present invention set numerous arithmetic unit on a chip, it can reduce hardware complexity and power consumption by shared arithmetic processing unit and shared register.一種基於線上遞回獨立成分分析之即時多通道腦波訊號處理器,其中,白化處理單元係將所接收之取樣訊號進行共變異數計算以產生共變異數矩陣,接著將該共變異數矩陣透過逆開方根矩陣計算單元之運算以產生白化矩陣,ORICA計算單元用於將該取樣訊號與該白化矩陣進行計算以得到白化後取樣訊號,並將該白化後取樣訊號與一解混合矩陣進行獨立成分分析運算以得到獨立成分資料,ORICA訓練單元係根據該獨立成份資料以對該解混合矩陣進行訓練以產生新的解混合矩陣,而新的解混合矩陣將提供ORICA計算單元於下一次獨立成分分析運算時使用。本發明將數個運算單元設置於一晶片上,透過共用暫存與共用算術運算單元的方法,以減少硬體複雜度及功耗。1‧‧‧基於線上遞回獨立成分分析之即時多通道腦波訊號處理器10‧‧‧逆開方根矩陣計算單元11‧‧‧白化處理單元12‧‧‧ORICA計算單元13‧‧‧ORICA訓練單元100‧‧‧取樣訊號101‧‧‧獨立成分資料