發(fā)布時間：2018-07-08 10:27

  本文選題：小波神經(jīng)網(wǎng)絡(luò) + 參數(shù)初始化��； 參考：《中北大學(xué)》2015年博士論文

【摘要】：小波分析與神經(jīng)網(wǎng)絡(luò)都是新一代計算智能信息處理技術(shù)的主要組成部分。小波變換是一個時間域和頻率域的局部變換,利用對小波函數(shù)的伸縮平移運(yùn)算對信號進(jìn)行不同尺度下的分析,可以有效地從信號中提取有用信息。它克服了傳統(tǒng)傅里葉變換不能同時進(jìn)行時頻分析的缺陷,因而成為了非線性科學(xué)的前沿技術(shù)之一。人工神經(jīng)網(wǎng)絡(luò)是通過對人腦神經(jīng)網(wǎng)絡(luò)的合理抽象而形成的理論化模型,它可以依靠自身對環(huán)境的自學(xué)習(xí)能力獲取知識,并利用神經(jīng)元之間的連接權(quán)值存儲獲取的知識。小波神經(jīng)網(wǎng)絡(luò)(簡稱小波網(wǎng)絡(luò))是小波分析與神經(jīng)網(wǎng)絡(luò)相結(jié)合的產(chǎn)物,它可以一方面保持神經(jīng)網(wǎng)絡(luò)的多輸入并行處理能力、自學(xué)習(xí)能力、非線性映射和容錯能力等,另一方面將具有強(qiáng)數(shù)學(xué)基礎(chǔ)的小波分析方便地應(yīng)用于高維問題,有效地發(fā)揮二者的優(yōu)勢。本文在研究了已有經(jīng)典小波神經(jīng)網(wǎng)絡(luò)模型的基礎(chǔ)上,以前饋式小波網(wǎng)絡(luò)為主要研究對象,對具有張量積型小波函數(shù)的自適應(yīng)小波網(wǎng)絡(luò)的初始化、與模糊機(jī)制結(jié)合的模糊小波網(wǎng)絡(luò)構(gòu)造與算法、徑向小波網(wǎng)絡(luò)的算法等方面做了較為深入的研究與分析,并將其應(yīng)用于信號預(yù)測,系統(tǒng)辨識,模式分類等方面,使得這幾類小波網(wǎng)絡(luò)在應(yīng)用方面具有更高的實(shí)用價值與意義。具體工作主要有:首先,具有張量積型小波激活函數(shù)的多維輸入自適應(yīng)小波網(wǎng)絡(luò)(WNN_M)是最早也是最經(jīng)典的小波網(wǎng)絡(luò)模型,其尺度參數(shù)、平移參數(shù)及線性權(quán)值均為網(wǎng)絡(luò)可調(diào)參數(shù),應(yīng)用中增加了網(wǎng)絡(luò)的靈活性,使得用更小規(guī)模的小波網(wǎng)絡(luò)逼近非線性系統(tǒng)成為可能,在其應(yīng)用中,初始參數(shù)的設(shè)置仍多采取Q.H.Zhang提出的啟發(fā)式方法(HIA)。為了得到針對該模型的更理想的參數(shù)初始值,使得網(wǎng)絡(luò)訓(xùn)練能夠快速收斂,提出了一種利用隱層小波函數(shù)時頻局部化特點(diǎn)的改進(jìn)參數(shù)初始化方法(CIA)。該方法首先根據(jù)多維輸入數(shù)據(jù)的統(tǒng)計特征設(shè)定閾值向量,利用最鄰近聚類的思想確定網(wǎng)絡(luò)隱層神經(jīng)元個數(shù),并由聚類后的類中心與半徑,通過隱層小波函數(shù)的時間窗寬確定相應(yīng)的小波平移參向量與尺度參向量的初始值。在對三類不同的時間序列預(yù)測的實(shí)驗(yàn)中看出,相比于WNN_M-HIA,WNN_M-CIA的初始誤差明顯較低,在經(jīng)過梯度下降法的數(shù)次迭代后,其預(yù)測精度仍優(yōu)于基于HIA初始化的網(wǎng)絡(luò)預(yù)測精度,說明了該初始化算法的有效性和合理性。其次,考慮到將模糊理論與小波分析、神經(jīng)網(wǎng)絡(luò)結(jié)合研究的趨勢,提出了一種將TSK模糊系統(tǒng)與小波網(wǎng)絡(luò)結(jié)合的模糊小波網(wǎng)絡(luò)模型。該模型將具有連續(xù)參數(shù)和張量積型多維小波激活函數(shù)的小波網(wǎng)絡(luò)作為TSK模糊系統(tǒng)的結(jié)論部分,分析了與已有模糊小波網(wǎng)絡(luò)模型的區(qū)別,并將其應(yīng)用于非線性系統(tǒng)辨識。在訓(xùn)練算法方面,選擇使用基于粒子群算法和在線梯度下降法結(jié)合的混合優(yōu)化算法。通過對兩個非線性系統(tǒng)進(jìn)行辨識實(shí)驗(yàn),可以看出,即使采用了較少的模糊規(guī)則和參數(shù)規(guī)模,相比于已有的模型,本文提出的模型與算法仍能得到更加滿意的辨識結(jié)果。再次,基于含有競爭算法的Kohonen自組織映射(SOM)神經(jīng)網(wǎng)絡(luò),提出了一種對車牌圖像進(jìn)行傾斜校正與字符分割的方法,將車牌中字符部分的像素點(diǎn)根據(jù)坐標(biāo)間的歐氏距離聚為七類,根據(jù)神經(jīng)元權(quán)值向量得到車牌的傾斜角度估計值,從而達(dá)到傾斜校正的目的。另外經(jīng)過預(yù)處理后,SOM算法得到的權(quán)值向量還可以進(jìn)一步利用最短距離法對車牌字符進(jìn)行分割,實(shí)例說明了本算法的有效性。最后,在SOM競爭算法的基礎(chǔ)上,結(jié)合徑向小波網(wǎng)絡(luò),提出了一種自生成圓盤細(xì)胞分裂算法,并作為前一工作的后續(xù),將其應(yīng)用于車牌字符的識別應(yīng)用中。該算法利用競爭機(jī)制將輸入模式映射到二維單位圓盤上而不是矩形域中的神經(jīng)元上,作為徑向小波神經(jīng)網(wǎng)絡(luò)的隱層神經(jīng)元,且不需事先給定規(guī)模,所需神經(jīng)元的個數(shù)與分布情況均可由網(wǎng)絡(luò)分類結(jié)果確定并由初始神經(jīng)元模仿細(xì)胞分裂的方式得到。其中用于賦值分裂后神經(jīng)元權(quán)值的“圓周近鄰策略”不僅可以使得輸入樣本的圓盤映射保持拓?fù)湫?還可以有效利用已經(jīng)通過競爭算法訓(xùn)練過的神經(jīng)元權(quán)值,使得算法更具高效性。通過對車牌中的英文字母樣本、英文字母或數(shù)字樣本進(jìn)行分類識別實(shí)驗(yàn),相比于經(jīng)典徑向基函數(shù)(RBF)網(wǎng)絡(luò),本文提出的算法可以在得到較小規(guī)模的網(wǎng)絡(luò)前提下達(dá)到更高的識別正確率,具有較高的實(shí)用價值。
[Abstract]:Wavelet analysis and neural network are the main components of the new generation of computing intelligent information processing technology. The wavelet transform is a local transformation of a time domain and frequency domain. Using the expansion and translation operation of the wavelet function to analyze the signal in different scales, the useful information can be extracted from the signal effectively. It overcomes the tradition. Fu Liye transform can not simultaneously carry on the defect of time frequency analysis, so it becomes one of the frontier technologies of nonlinear science. The artificial neural network is a theoretical model formed by the rational abstraction of the neural network of the human brain. It can rely on self learning ability of the environment to obtain knowledge, and use the connection weights between the neurons. The wavelet neural network (WNN) is a product of the combination of wavelet analysis and neural network. It can maintain the ability of multi input parallel processing, self learning, nonlinear mapping and fault tolerance on one hand. On the other hand, the wavelet analysis with strong mathematical basis is easily applied to the high level. On the basis of the existing classical wavelet neural network model, this paper has studied the existing classical wavelet neural network model. The former is the main research object, the initialization of the adaptive wavelet network with tensor product type wavelet function, the structure and algorithm of the fuzzy wavelet network combining with the fuzzy mechanism, the radial wavelet network. The algorithm of the collaterals has been studied and analyzed in detail, and it is applied to signal prediction, system identification, pattern classification and so on, which make the wavelet networks have higher practical value and significance in application. The main work is: first, multidimensional input adaptive wavelet with the tensor product type wavelet activation function Network (WNN_M) is the earliest and most classic wavelet network model. Its scale parameters, translation parameters and linear weights are all adjustable parameters of the network. In application, the flexibility of the network is added, making it possible to approximate the nonlinear system with a smaller scale wavelet network. In its application, the initial parameters are still more Q.H.Zhang proposed. The heuristic method (HIA). In order to get the more ideal initial parameters for the model and make the network training fast convergence, an improved parameter initialization method (CIA) is proposed using the time-frequency localization feature of the hidden layer wavelet function (CIA). The method first sets the threshold vector according to the statistical characteristics of the multidimensional input data, and uses the most important method. The idea of adjacent clustering determines the number of neurons in the hidden layer of the network, and determines the initial value of the corresponding wavelet translation reference vector and the scale parameter by the time window width of the hidden layer wavelet function. The initial error of the three different time series prediction shows that the initial error of the WNN_M-CIA is compared to the WNN_M-HIA. It is obviously lower. After several iterations of the gradient descent method, the prediction accuracy is still better than the network prediction accuracy based on HIA initialization. It shows the validity and rationality of the initialization algorithm. Secondly, considering the trend of combining the fuzzy theory with the wavelet analysis and the neural network, a kind of TSK fuzzy system and the wavelet network are proposed. This model uses the wavelet network with continuous parameter and tensor product multidimensional wavelet activation function as the conclusion part of the TSK fuzzy system, analyzes the difference between the fuzzy wavelet network model and the existing fuzzy wavelet network model, and applies it to the nonlinear system identification. In the training algorithm, the selection of the particle swarm optimization is based on the particle swarm optimization. The hybrid optimization algorithm combined with the online gradient descent method. Through the identification experiments of two nonlinear systems, it can be seen that even if less fuzzy rules and parameters are used, compared with the existing models, the proposed model and algorithm can still get more satisfactory identification results. Again, based on the competition algorithm. The Kohonen self organizing mapping (SOM) neural network has proposed a method of tilt correction and character segmentation for the license plate images. The pixels of the characters in the license plate are divided into seven classes according to the Euclidean distance between the coordinates, and the estimation of the tilt angle of the license plate is obtained according to the weight vector of the neuron. In addition, the purpose of the tilt correction is achieved. After preprocessing, the weight vector obtained by the SOM algorithm can also further use the shortest distance method to segment the character of the license plate. The example shows the effectiveness of the algorithm. Finally, on the basis of the SOM competition algorithm, a self generated circular disk cell division algorithm is proposed with the radial wavelet network, which will be used as a follow-up of the previous work. It is applied to the recognition of license plate characters. The algorithm uses competition mechanism to map the input mode to the two-dimensional unit disk instead of the neuron in the rectangular domain. As a hidden layer neuron of the radial wavelet neural network, the number and distribution of the required neurons can be classified by the network classification results. The "circumference nearest neighbor strategy", which is used to assign the weights of the neurons after the splitting, can not only make the disk mapping of the input sample keep the topological order, but also effectively use the weights that have been trained by the competition algorithm, making the algorithm more efficient. Compared with the classic radial basis function (RBF) network, the proposed algorithm can achieve higher recognition accuracy and have higher practical value than the classical radial basis function (RBF) network.
【學(xué)位授予單位】：中北大學(xué)
【學(xué)位級別】：博士
【學(xué)位授予年份】：2015
【分類號】：TP183;TP391.4

【相似文獻(xiàn)】

相關(guān)期刊論文 前10條

1 呂朝霞,胡維禮;小波網(wǎng)絡(luò)在控制系統(tǒng)中的應(yīng)用[J];信息與控制;2000年06期

2 呂立華,宋執(zhí)環(huán),李平;一種小波網(wǎng)絡(luò)設(shè)計新方法[J];信息與控制;2002年01期

3 劉志剛,王曉茹,錢清泉;小波網(wǎng)絡(luò)的研究進(jìn)展與應(yīng)用[J];電力系統(tǒng)自動化;2003年06期

4 彭觀明;李志強(qiáng);劉照軍;;用小波網(wǎng)絡(luò)法預(yù)測提升機(jī)故障趨勢[J];煤礦機(jī)械;2007年02期

5 俞保;朱家海;;基于小波網(wǎng)絡(luò)的混沌時間序列預(yù)測[J];科技資訊;2007年08期

6 劉彤彤;;基于小波網(wǎng)絡(luò)的復(fù)雜系統(tǒng)建模方法研究[J];科技咨詢導(dǎo)報;2007年11期

7 王軍;肖洪梅;武卯泉;丁傳東;盧杰;;一種模糊小波網(wǎng)絡(luò)逼近性分析(英文)[J];西華大學(xué)學(xué)報(自然科學(xué)版);2007年06期

8 郭張軍;宋漢周;彭鵬;;小波網(wǎng)絡(luò)模型在滲流量預(yù)測中的應(yīng)用[J];勘察科學(xué)技術(shù);2008年01期

9 陳濤,屈梁生;多分辨小波網(wǎng)絡(luò)的理論及應(yīng)用[J];中國機(jī)械工程;1997年02期

10 張西寧;連續(xù)參數(shù)小波網(wǎng)絡(luò)在機(jī)械狀態(tài)預(yù)報中的應(yīng)用[J];西安交通大學(xué)學(xué)報;1999年03期

相關(guān)會議論文 前10條

1 施陽;徐德民;任章;;基于小波網(wǎng)絡(luò)的智能PID控制[A];1996年中國控制會議論文集[C];1996年

2 徐長江;;小波網(wǎng)絡(luò)[A];1997年中國控制會議論文集[C];1997年

3 徐志敏;;基于改進(jìn)粒子群算法的小波網(wǎng)絡(luò)及其應(yīng)用[A];第三屆中國智能計算大會論文集[C];2009年

4 劉華;趙薇;李占峰;;小波網(wǎng)絡(luò)在直接轉(zhuǎn)矩控制系統(tǒng)中的應(yīng)用研究[A];2009中國控制與決策會議論文集（3）[C];2009年

5 張春紅;羅洪霞;王媛;李天云;;基于徑向基小波網(wǎng)絡(luò)的高壓輸電線短路故障選相[A];第十屆全國電工數(shù)學(xué)學(xué)術(shù)年會論文集[C];2005年

6 黃德先;金以慧;張杰;MORRISJuhan;;基于小波網(wǎng)絡(luò)模型的EPI過程的控制和優(yōu)化[A];第16屆中國過程控制學(xué)術(shù)年會暨第4屆全國故障診斷與安全性學(xué)術(shù)會議論文集[C];2005年

7 葉中行;陳凱風(fēng);;基于粗糙集和小波網(wǎng)絡(luò)的銀行信貸信用分類方法[A];中國現(xiàn)場統(tǒng)計研究會第十三屆學(xué)術(shù)年會論文集[C];2007年

8 項(xiàng)海林;張殿治;;用小波網(wǎng)絡(luò)進(jìn)行非線性逼近[A];圖像 仿真 信息技術(shù)——第二屆聯(lián)合學(xué)術(shù)會議論文集[C];2002年

9 張新紅;;基于小波網(wǎng)絡(luò)的經(jīng)濟(jì)時間序列預(yù)測模型[A];2002年中國管理科學(xué)學(xué)術(shù)會議論文集[C];2002年

10 黃敏;朱啟兵;崔寶同;;基于小波網(wǎng)絡(luò)的板形板厚系統(tǒng)PID神經(jīng)網(wǎng)絡(luò)控制[A];第二十七屆中國控制會議論文集[C];2008年

相關(guān)博士學(xué)位論文 前7條

1 程蓉;基于小波的神經(jīng)網(wǎng)絡(luò)研究及其在信息處理中的應(yīng)用[D];中北大學(xué);2015年

2 呂立華;復(fù)雜工業(yè)系統(tǒng)基于小波網(wǎng)絡(luò)與魯棒估計建模方法研究[D];浙江大學(xué);2001年

3 呂偉杰;小波網(wǎng)絡(luò)與多模型自適應(yīng)控制理論的研究[D];天津大學(xué);2003年

4 馮登超;蟻群算法與小波網(wǎng)絡(luò)在復(fù)雜性科學(xué)中的應(yīng)用研究[D];天津大學(xué);2008年

5 呂淑萍;小波網(wǎng)絡(luò)建模預(yù)報方法研究及其在股市預(yù)測中的應(yīng)用[D];哈爾濱工程大學(xué);2004年

6 王軍;模糊小波網(wǎng)絡(luò)及其在永磁同步電機(jī)控制中的應(yīng)用[D];西南交通大學(xué);2005年

7 吳清;腦電信號源數(shù)值解法的研究[D];河北工業(yè)大學(xué);2001年

相關(guān)碩士學(xué)位論文 前10條

1 劉元印;基于改進(jìn)小波網(wǎng)絡(luò)算法的施工期橋梁船撞風(fēng)險評估模型[D];哈爾濱工業(yè)大學(xué);2015年

2 趙超;一種小波網(wǎng)絡(luò)的快速算法及其在控制中的應(yīng)用[D];大連理工大學(xué);2005年

3 徐志敏;小波網(wǎng)絡(luò)的研究及其應(yīng)用[D];山東師范大學(xué);2005年

4 郭俊元;一種改進(jìn)小波網(wǎng)絡(luò)及其應(yīng)用[D];吉林大學(xué);2006年

5 李萍;小波網(wǎng)絡(luò)在經(jīng)濟(jì)預(yù)測中的應(yīng)用[D];浙江大學(xué);2010年

6 蔣偉;小波網(wǎng)絡(luò)的相關(guān)問題研究[D];電子科技大學(xué);2013年

7 閆璐;小波網(wǎng)絡(luò)在非線性控制系統(tǒng)中的應(yīng)用研究[D];華北電力大學(xué);2013年

8 牛國鵬;小波網(wǎng)絡(luò)與混沌時間序列預(yù)測[D];蘭州大學(xué);2009年

9 潘俊林;基于小波網(wǎng)絡(luò)的沖激信號檢測方法研究[D];五邑大學(xué);2008年

10 肖廣云;利用小波網(wǎng)絡(luò)處理凝析天然氣流量計測量數(shù)據(jù)[D];中國石油大學(xué);2008年

，

本文編號：2107134