系统工程与电子技术
繫統工程與電子技術
계통공정여전자기술
SYSTEMS ENGINEERING AND ELECTRONICS
2014年
4期
619-623
,共5页
付卫红%李爱丽%马丽芬%黄坤%严新
付衛紅%李愛麗%馬麗芬%黃坤%嚴新
부위홍%리애려%마려분%황곤%엄신
欠定盲源分离%混合矩阵估计%势函数%基准值%尺度参数
欠定盲源分離%混閤矩陣估計%勢函數%基準值%呎度參數
흠정맹원분리%혼합구진고계%세함수%기준치%척도삼수
underdetermined blind separation%mixing matrix estimation%potential function%reference va-lue%scale parameter
针对 k 维子空间法尺度参数选择的盲目性,提出了基于估计参数的势函数方法。首先根据观测信号估计出基准值 r 1,并根据基准值 r 1选取势函数 f 的尺度参数σ1,由势函数 f 估计出聚类平面后,根据聚类平面估计出基准值 r2,根据基准值 r2选取势函数 g 的尺度参数σ2,并由势函数 g 估计出混合矩阵。该方法充分利用了观测信号的统计特性,实验结果表明,在无噪条件下,该方法比改进前的方法在矩阵估计方面误差减小了75%,较其他方法误差减小了1~2个数量级;在16 dB 信噪比下,该方法比改进前在矩阵估计精度上提高8.7%,源信号个数估计准确率提高了1倍。
針對 k 維子空間法呎度參數選擇的盲目性,提齣瞭基于估計參數的勢函數方法。首先根據觀測信號估計齣基準值 r 1,併根據基準值 r 1選取勢函數 f 的呎度參數σ1,由勢函數 f 估計齣聚類平麵後,根據聚類平麵估計齣基準值 r2,根據基準值 r2選取勢函數 g 的呎度參數σ2,併由勢函數 g 估計齣混閤矩陣。該方法充分利用瞭觀測信號的統計特性,實驗結果錶明,在無譟條件下,該方法比改進前的方法在矩陣估計方麵誤差減小瞭75%,較其他方法誤差減小瞭1~2箇數量級;在16 dB 信譟比下,該方法比改進前在矩陣估計精度上提高8.7%,源信號箇數估計準確率提高瞭1倍。
침대 k 유자공간법척도삼수선택적맹목성,제출료기우고계삼수적세함수방법。수선근거관측신호고계출기준치 r 1,병근거기준치 r 1선취세함수 f 적척도삼수σ1,유세함수 f 고계출취류평면후,근거취류평면고계출기준치 r2,근거기준치 r2선취세함수 g 적척도삼수σ2,병유세함수 g 고계출혼합구진。해방법충분이용료관측신호적통계특성,실험결과표명,재무조조건하,해방법비개진전적방법재구진고계방면오차감소료75%,교기타방법오차감소료1~2개수량급;재16 dB 신조비하,해방법비개진전재구진고계정도상제고8.7%,원신호개수고계준학솔제고료1배。
A potential function method with estimated parameter is proposed for blindness of scale parame-ter selecting of k-dimensional subspace method.Firstly,the reference value r 1 is estimated based on observed signals and the scale parameterσ1 for potential function f is estimated based on r 1 .When cluster planes are esti-mated by f ,the reference value r 2 is estimated based on them and the scale parameterσ2 for potential function g is estimated based on r 2 .Thus,we obtain the estimation of mixing matrix.This method makes full use of the statistical properties of observed signals.The experiment shows that this method is reduced by 75% in matrix estimation error than the method before improved and is reduced by 1 2 orders of magnitude compared with other methods when there is no noise;the estimated precision is 8.7% higher and the accuracy rate of the source’s number estimation is twice higher.