一种基于单位超球面平均的盲反卷积方法.pdfVIP

绗?5 鍗 绗? 鏈 鎺 鍒 涓 鍐 绛 2010 骞 5 鏈 Vol. 25 No. 5 Control and Decision May 2010 鏂囩珷缂栧彿: 1001-0920 (2010) 05-0758-05 涓€绉嶅熀浜庡崟浣嶈秴鐞冮潰骞冲潎鐨勭洸鍙嶅嵎绉柟娉 绋 鎴? 钀ц暣璇? 宀崇户鍏 ( 鍚屾祹澶у鐢靛瓙涓庝俊鎭伐绋嬪闄紝涓婃捣201804) 鎽 瑕侊細锛氾細鍦ㄦ満鍣ㄥ涔犵悊璁轰腑, 涓烘秷闄ら殢鏈哄奖鍝? 瀵规牱鏈繘琛屽钩鍧囨槸涓€绉嶅箍涓轰娇鐢ㄧ殑澶勭悊鎵嬫. 涓烘, 鍦ㄥ叿鏈夌浉鍚岀墿鐞 婵€鍙戣繃绋嬬殑绯荤粺涓? 鍒╃敤榛庢浖姊害鐨勬蹇? 鏋勯€犱簡鍗曚綅瓒呯悆闈笂鐨勫钩鍧囩畻瀛? 浠ュ湴闇囨暟鎹洸鍙嶅嵎绉负渚? 缁撳悎鏀硅繘鍨 鐨勫甫鐘剁嫭绔嬪垎閲忓垎鏋? 鎻愬嚭浜嗗熀浜庡崟浣嶈秴鐞冮潰骞冲潎鐨勭洸鍙嶅嵎绉柟娉? 浠跨湡瀹為獙琛ㄦ槑浜嗘墍鎻愬嚭鏂规硶鐨勬湁鏁堟€у拰浼樿秺鎬? 鍏抽敭璇嶏細锛氾細甯︾姸鐙珛鍒嗛噺鍒嗘瀽锛涚洸鍙嶅嵎绉紱鍗曚綅瓒呯悆闈紱榛庢浖姊害 涓浘鍒嗙被鍙凤細锛氾細TN911; O186 鏂囩尞鏍囪瘑鐮侊細锛氾細A Blind deconvolution method based on averaging over unit hypersphere CHENG Cheng, XIAO Yun-shi , YUE Ji-guang (School of Electronics and Information Engineering锛孴ongji University 锛孲hanghai 201804锛孋hina. Correspondent: CHENG Cheng锛孍-mail 锛歝hengcheng lcc@163.com) Abstract: Sample averaging is a widely used method in machine learning to rid the random in铿倁ence. Therefore, in the system with the same physical process, an averaging operator on the unit hypersphere is constructed with the idea of Riemannian gradient. Combined with the improved banded independent component analysis in the problem of blind deconvolution for seismic data, a blind deconvolution method based on averaging over unit hypersphere is proposed. Simulation result shows the effectiveness and superiority of the method. Key words: Banded independent component analysis 锛汢lind deconvolution 锛沀nit hypersphere 锛汻iemannian gradient 1 寮曞紩寮 瑷€瑷€瑷€ 瀹為檯鍦伴渿鍕樻帰涓? 涓€鑸涓洪偦杩戦亾闂寸殑鍦伴渿鏁版嵁鍏锋湁 鐩插弽鍗风Н浣滀负涓€绉嶅吀鍨嬬殑鍙嶉棶棰? 鍦ㄤ俊閬撳潎琛°€ 寰堝己鐨勭浉鍏虫€? 鑰孊ICA 鍙嶅嵎绉殑鍗曢亾澶勭悊浼氱粰闂 [1] 澧炲姞闅忔満涓嶇‘瀹氭€? 绯荤粺杈ㄨ瘑銆佽嚜閫傚簲婊ゆ尝绛夐鍩熶竴鐩村鍙