• PUDN用户
    了解作者
  • PDF
    开发工具
  • 234KB
    文件大小
  • rar
    文件格式
  • 0
    收藏次数
  • 1 积分
    下载积分
  • 12
    下载次数
  • 2012-04-25 09:28
    上传日期
这是一篇论文,说的是关于哼唱检索系统中的基频检测,用的算法是短时平均幅度差函数
Pitch-Detection.rar
  • Pitch Detection.pdf
    264.6KB
内容介绍
<html xmlns="http://www.w3.org/1999/xhtml"> <head> <meta charset="utf-8"> <meta name="generator" content="pdf2htmlEX"> <meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1"> <link rel="stylesheet" href="https://static.pudn.com/base/css/base.min.css"> <link rel="stylesheet" href="https://static.pudn.com/base/css/fancy.min.css"> <link rel="stylesheet" href="https://static.pudn.com/prod/directory_preview_static/622b6f8515da9b288bd74e63/raw.css"> <script src="https://static.pudn.com/base/js/compatibility.min.js"></script> <script src="https://static.pudn.com/base/js/pdf2htmlEX.min.js"></script> <script> try{ pdf2htmlEX.defaultViewer = new pdf2htmlEX.Viewer({}); }catch(e){} </script> <title></title> </head> <body> <div id="sidebar" style="display: none"> <div id="outline"> </div> </div> <div id="pf1" class="pf w0 h0" data-page-no="1"><div class="pc pc1 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/622b6f8515da9b288bd74e63/bg1.jpg"><div class="t m0 x1 h2 y1 ff1 fs0 fc0 sc0 ls0 ws0">The Ninth International C<span class="_ _0"></span>onference on Electronic Measurem<span class="_ _0"></span>ent &amp; Instruments <span class="_"> </span><span class="ls1 ws1"> <span class="_ _1"></span> ICEMI&#8217;2009 </span></div><div class="t m0 x2 h3 y2 ff2 fs1 fc0 sc0 ls2 ws2">Improving AMDF for Pitch Period Detection </div><div class="t m0 x3 h4 y3 ff1 fs2 fc0 sc0 ls3 ws3">Kang Guangyu</div><div class="t m0 x4 h5 y4 ff1 fs3 fc0 sc0 ls4 ws4">1</div><div class="t m0 x5 h4 y3 ff1 fs2 fc0 sc0 ls4 ws4"> Guo Shize</div><div class="t m0 x6 h5 y4 ff1 fs3 fc0 sc0 ls4 ws4">2</div><div class="t m0 x7 h2 y5 ff1 fs0 fc0 sc0 ls5 ws5">1 Department of <span class="_ _2"></span>Automatic <span class="_ _0"></span>T<span class="_ _2"></span>est and <span class="ls6 ws6">Control, Harbin Institute of T<span class="_ _2"></span>echnol<span class="_ _0"></span><span class="ls7 ws7">ogy<span class="_ _2"></span>, Harbin, Heil<span class="ls8 ws8">ongjiang 150001 </span></span></span></div><div class="t m0 x8 h2 y6 ff1 fs0 fc0 sc0 ls9 ws9">2 No.54 Institute from Headqu<span class="lsa wsa">arters of the General Staff<span class="_ _3"> </span><span class="lsb wsb">Beijing 100001 </span></span></div><div class="t m0 x9 h2 y7 ff1 fs0 fc0 sc0 ls6 ws6">E-mail: lijunbao@hit.edu.cn </div><div class="t m0 x1 h6 y8 ff3 fs4 fc0 sc0 lsc ws4">Abstract: <span class="_"> </span><span class="ff4 lsd wsc">Pitch period is the important parameters of speech </span></div><div class="t m0 x1 h6 y9 ff4 fs4 fc0 sc0 lse wsd">recognition and speech synthesis<span class="_ _0"></span>. Pitch period detection has b<span class="_ _0"></span>een </div><div class="t m0 x1 h6 ya ff4 fs4 fc0 sc0 lsf wse">focus in the field of audio processing research. Traditional<span class="_ _0"></span> </div><div class="t m0 x1 h6 yb ff4 fs4 fc0 sc0 ls10 wsf">AMDF-based algorithm and its improved version, </div><div class="t m0 x1 h6 yc ff4 fs4 fc0 sc0 ls11 ws10">LV-AMDF-based algorithm easily leads to the double error or </div><div class="t m0 x1 h6 yd ff4 fs4 fc0 sc0 ls12 ws11">half error, and so on in the pitch detection. To solve thes<span class="_ _0"></span>e </div><div class="t m0 x1 h6 ye ff4 fs4 fc0 sc0 ls13 ws12">problems, AMDF, LV-AMDF function cha<span class="_ _0"></span>racteristics and </div><div class="t m0 x1 h6 yf ff4 fs4 fc0 sc0 lse ws13">shortcomings in pitch detection are analyzed<span class="_ _0"></span>, the parameters </div><div class="t m0 x1 h6 y10 ff4 fs4 fc0 sc0 ls14 ws14">compensation AMDF pitch detection algorithm is<span class="_ _0"></span> proposed in </div><div class="t m0 x1 h6 y11 ff4 fs4 fc0 sc0 lsc ws15">this article to reduce semi-frequency, double-frequency <span class="_ _0"></span>errors </div><div class="t m0 x1 h6 y12 ff4 fs4 fc0 sc0 ls15 ws16">often appear in the pitch detection and impro<span class="_ _0"></span>ve the detection </div><div class="t m0 x1 h6 y13 ff4 fs4 fc0 sc0 ls14 ws17">accuracy. Experimental results show that its pitch detection </div><div class="t m0 x1 h6 y14 ff4 fs4 fc0 sc0 ls16 ws18">accuracy is better than AMDF and LV-AMDF.</div><div class="t m0 x1 h6 y15 ff3 fs4 fc0 sc0 ls14 ws4">Keywords:<span class="_"> </span><span class="ff4 lsc ws19">Pitch detection algorithm; AMDF; Length-Varied </span></div><div class="t m0 x1 h6 y16 ff4 fs4 fc0 sc0 ls17 ws1a">AMDF; amplitude compensation </div><div class="t m1 xa h7 y17 ff5 fs5 fc0 sc0 ls4 ws4">&#727;</div><div class="t m0 xb h2 y18 ff1 fs0 fc0 sc0 ls18 ws4">I.<span class="_ _4"> </span>INTRODUCTION </div><div class="t m0 xc h2 y19 ff1 fs0 fc0 sc0 ls19 ws1b">Speech signals can be divided into voiced and voiceless</div><div class="t m0 x1 h2 y1a ff1 fs0 fc0 sc0 ls1a ws4">by<span class="ls1b ws1c"> the vibration of vocal cord, and <span class="_ _0"></span>voiced is with a </span></div><div class="t m0 x1 h2 y1b ff1 fs0 fc0 sc0 ls1c ws1d">quasi-cyclical. As an important characteristic parameter, </div><div class="t m0 x1 h2 y1c ff1 fs0 fc0 sc0 ls1c ws1e">pitch period is widely applied in speech recognition, </div><div class="t m0 x1 h2 y1d ff1 fs0 fc0 sc0 ls1d ws1f">speaker recognition, speech synthesis, and other research </div><div class="t m0 x1 h2 y1e ff1 fs0 fc0 sc0 ls1e ws20">fields [1,2]. So the detection me<span class="ls1b ws21">thod of <span class="_ _0"></span>pitch period is focus </span></div><div class="t m0 x1 h2 y1f ff1 fs0 fc0 sc0 ls1d ws22">in the research field of speech signal processing. On the </div><div class="t m0 x1 h2 y20 ff1 fs0 fc0 sc0 ls19 ws23">other hand, since there&#8217;s large difference and instability in </div><div class="t m0 x1 h2 y21 ff1 fs0 fc0 sc0 ls19 ws24">characteristics of the channel, ch<span class="ws25">annel resonance peak will </span></div><div class="t m0 x1 h2 y22 ff1 fs0 fc0 sc0 ls1f ws26">seriously impact on the harm<span class="_ _0"></span>onic structure of the speech </div><div class="t m0 x1 h2 y23 ff1 fs0 fc0 sc0 ls1f ws27">signal, and the pitch period is <span class="ls20 ws28">of the very wide range, plus is </span></div><div class="t m0 x1 h2 y24 ff1 fs0 fc0 sc0 ls1b ws29">vulnerable to tone and emotion, <span class="ls20 ws2a">and the other factors, so </span></div><div class="t m0 x1 h2 y25 ff1 fs0 fc0 sc0 ls21 ws2b">pitch period quickly and effec<span class="ls20 ws2c">tively detection is a very </span></div><div class="t m0 x1 h2 y26 ff1 fs0 fc0 sc0 ls22 ws2d">complex issue. Against the qu<span class="ws2e">asi-periodic si<span class="_ _0"></span>gnal of voiced, </span></div><div class="t m0 x1 h2 y27 ff1 fs0 fc0 sc0 ls23 ws2f">usually short-time technology is used, the most com<span class="_ _0"></span>monly </div><div class="t m0 x1 h2 y28 ff1 fs0 fc0 sc0 ls24 ws30">used method of short-t<span class="ls25 ws31">im<span class="_ _0"></span>e estimation of pitch period is: </span></div><div class="t m0 x1 h2 y29 ff1 fs0 fc0 sc0 ls20 ws28">Auto-correlation and AMDF pitch detection algorithm<span class="_ _0"></span>[1-3]. </div><div class="t m0 xc h2 y2a ff1 fs0 fc0 sc0 ls25 ws32">Since the advantage of auto<span class="ls1c ws33">-correlation method is a </span></div><div class="t m0 x1 h2 y2b ff1 fs0 fc0 sc0 ls26 ws4">simp<span class="ls23 ws34">le algorithm and the extracted pitch period is more </span></div><div class="t m0 x1 h2 y2c ff1 fs0 fc0 sc0 ls22 ws27">accurate, but usually the center of the speech signal needs to </div><div class="t m0 x1 h2 y2d ff1 fs0 fc0 sc0 ls27 ws35">be preprocessed with clipping. The choice of clipping </div><div class="t m0 x1 h2 y2e ff1 fs0 fc0 sc0 ls27 ws36">threshold level is often difficult <span class="ls28 ws37">to determine. If the choice </span></div><div class="t m0 x1 h2 y2f ff1 fs0 fc0 sc0 ls29 ws38">is determined incorrectly, a larg<span class="ws39">e deviation will appear in </span></div><div class="t m0 x1 h2 y30 ff1 fs0 fc0 sc0 ls1c ws3a">estimation of pitch period. <span class="ls23 ws3b">And the method is ve<span class="_ _0"></span>ry </span></div><div class="t m0 x1 h2 y31 ff1 fs0 fc0 sc0 ls2a ws3c">influenced by the noise with <span class="ls1f ws3d">less robustness performance </span></div><div class="t m0 x1 h2 y32 ff1 fs0 fc0 sc0 ls1f ws3e">against noise. AMDF method has<span class="_ _0"></span> advantages such as sm<span class="_ _0"></span>all </div><div class="t m0 x1 h2 y33 ff1 fs0 fc0 sc0 ls1f ws3f">computing amount, high <span class="_ _0"></span>preci<span class="ls2b ws40">sion, and so on.<span class="_ _0"></span> It has been </span></div><div class="t m0 x1 h2 y34 ff1 fs0 fc0 sc0 ls23 ws41">widely used in the field of speech coding. Experiments </div><div class="t m0 x1 h2 y35 ff1 fs0 fc0 sc0 ls22 ws42">show that when the speech signal changes flatly, the </div><div class="t m0 x1 h2 y36 ff1 fs0 fc0 sc0 ls1f ws27">estimation of pitch period using <span class="_"> </span> </div><div class="t m0 xd h2 y37 ff1 fs0 fc0 sc0 ls1f ws27">AMDF can get higher estimation accuracy and reliability. </div><div class="t m0 x1 h2 y38 ff1 fs0 fc0 sc0 ls29 ws43">And the calculation cost needed is less th<span class="ls21 ws44">an that of </span></div><div class="t m0 x1 h2 y39 ff1 fs0 fc0 sc0 ls2c ws45">auto-correlation. But when th<span class="ls27 ws35">e magnitude or the pitch </span></div><div class="t m0 x1 h2 y3a ff1 fs0 fc0 sc0 ls21 ws46">period of speech signal changes rapidly, AMDF method <span class="_"> </span> </div><div class="t m0 xe h2 y3b ff1 fs0 fc0 sc0 ls2d ws47">will decreased apparently in<span class="ls29 ws48"> pitch estimation </span></div><div class="t m0 xe h2 y3c ff1 fs0 fc0 sc0 ls1d ws49">accuracy, which directly affects the recognition </div><div class="t m0 xe h2 y3d ff1 fs0 fc0 sc0 ls19 ws4a">accuracy of Ch<span class="ls1c ws4b">inese tones[6]. Such as frequently the </span></div><div class="t m0 xe h2 y3e ff1 fs0 fc0 sc0 ls29 ws4c">case appears that, the pitch is estimated as the two </div><div class="t m0 xe h2 y3f ff1 fs0 fc0 sc0 ls29 ws4d">times or half of the actual <span class="ls2e ws25">results of the pitch period </span></div><div class="t m0 xe h2 y40 ff1 fs0 fc0 sc0 ls2a ws4e">[1]. Against the problem<span class="ls23 ws4f">s related documents have </span></div><div class="t m0 xe h2 y41 ff1 fs0 fc0 sc0 ls21 ws2b">proposed a lot of ways to<span class="ls2b ws50"> improve [1,2,3,8]. The </span></div><div class="t m0 xe h2 y42 ff1 fs0 fc0 sc0 ls22 ws51">literature [6] put forward (Length-Varied AMDF) </div><div class="t m0 xe h2 y43 ff1 fs0 fc0 sc0 ls2a ws52">method, pitch detection <span class="ls2b ws53">errors caused by the rapidly </span></div><div class="t m0 xe h2 y44 ff1 fs0 fc0 sc0 ls23 ws54">change of the frequency was corrected. However, </div><div class="t m0 xe h2 y45 ff1 fs0 fc0 sc0 ls1d ws55">there are some problems such as the value of the pitch </div><div class="t m0 xe h2 y46 ff1 fs0 fc0 sc0 ls29 ws56">period estimation matches ma<span class="ls2f ws57">inly the front part of the </span></div><div class="t m0 xe h2 y47 ff1 fs0 fc0 sc0 ls21 ws58">speech frame, and the bias on the back part is </div><div class="t m0 xe h2 y48 ff1 fs0 fc0 sc0 ls2b ws59">growing. Voicing mixed the frame (or the dullness of </div><div class="t m0 xe h2 y49 ff1 fs0 fc0 sc0 ls28 ws5a">the initial frame) has large deviations in computing </div><div class="t m0 xe h2 y4a ff1 fs0 fc0 sc0 ls1c ws4">results. </div><div class="t m0 xf h2 y4b ff1 fs0 fc0 sc0 ls1d ws5b">Speech signal is a time-variable unstable signal </div><div class="t m0 xe h2 y4c ff1 fs0 fc0 sc0 ls30 ws5c">with its a<span class="ls21 ws5d">mplitude attenuation by time, the pitch </span></div><div class="t m0 xe h2 y4d ff1 fs0 fc0 sc0 ls20 ws5e">period waves near the average of the pitch period. </div><div class="t m0 xe h2 y4e ff1 fs0 fc0 sc0 ls1b ws5f">Based on the phenom<span class="_ _0"></span>enon, the amplitude <span class="_ _0"></span>parameter </div><div class="t m0 xe h2 y4f ff1 fs0 fc0 sc0 ls19 ws60">compensation AMDF is proposed in this <span class="_ _0"></span>article to </div><div class="t m0 xe h2 y50 ff1 fs0 fc0 sc0 ls2a ws61">reduce pitch period detection errors caused by </div><div class="t m0 xe h2 y51 ff1 fs0 fc0 sc0 ls1e ws62">amplitude attenuation. Expe<span class="_ _0"></span>ri<span class="ls2b ws63">mental results show that </span></div><div class="t m0 xe h2 y52 ff1 fs0 fc0 sc0 ls28 ws64">its pitch detection accuracy is better than AMDF and </div><div class="t m0 xe h2 y53 ff1 fs0 fc0 sc0 ls1c ws4">LV-AMDF.</div><div class="t m0 x10 h2 y54 ff1 fs0 fc0 sc0 ls31 ws4">II.<span class="_ _5"> </span>ALGORITHM</div><div class="t m0 x11 h2 y55 ff1 fs0 fc0 sc0 ls8 ws4">COMPARISON</div><div class="t m0 xf h8 y56 ff1 fs0 fc0 sc0 ls2a ws65">AMDF pitch detection algorithm is defined as<span class="_ _0"></span><span class="ff6 ls4 ws4">&#14822;</span></div><div class="t m2 x12 h9 y57 ff7 fs6 fc0 sc0 ls32 ws4">&#58882;&#58883;<span class="_ _6"> </span>&#58882;<span class="_ _7"> </span>&#58883;<span class="_ _8"> </span>&#58882;&#58883;</div><div class="t m0 x13 ha y58 ff1 fs7 fc0 sc0 ls4 ws4">1</div><div class="t m0 x14 ha y59 ff1 fs7 fc0 sc0 ls4 ws4">0</div><div class="t m0 x15 hb y58 ff4 fs7 fc0 sc0 ls33 ws4">Nk</div><div class="t m0 x16 hb y5a ff4 fs7 fc0 sc0 ls34 ws4">ww</div><div class="t m0 x17 hb y59 ff4 fs7 fc0 sc0 ls4 ws4">n</div><div class="c x18 y5b w2 hc"><div class="t m0 x0 hd y5c ff4 fs8 fc0 sc0 ls4 ws4">D</div></div><div class="t m0 x19 hd y5d ff4 fs8 fc0 sc0 ls35 ws4">ks<span class="_ _9"></span>n<span class="_ _a"></span>k<span class="_ _b"></span>s</div><div class="t m0 x1a he y58 ff7 fs7 fc0 sc0 ls36 ws4">&#58884;&#58884;</div><div class="t m0 x1b he y59 ff7 fs7 fc0 sc0 ls4 ws4"> </div><div class="t m0 x1c hf y5d ff7 fs8 fc0 sc0 ls37 ws4"> &#58886;<span class="_ _c"></span>&#58884;</div><div class="t m0 x17 h10 y5e ff7 fs9 fc0 sc0 ls4 ws4">&#166;</div><div class="t m0 x1d hd y5d ff4 fs8 fc0 sc0 ls4 ws4">n</div><div class="t m0 x1e h2 y5f ff1 fs0 fc0 sc0 ls38 ws4"> (<span class="_ _d"></span>1<span class="_ _d"></span>)<span class="_ _e"></span> </div><div class="t m3 x1f h11 y60 ff7 fsa fc0 sc0 ls39 ws4">&#58882;&#58883;<span class="_ _f"> </span>&#58882;&#58883;<span class="_ _10"> </span>&#58882;&#58883;</div><div class="t m0 x20 h12 y61 ff4 fsb fc0 sc0 ls4 ws4">w</div><div class="c x21 y62 w3 h13"><div class="t m0 x0 h14 y63 ff4 fsc fc0 sc0 ls4 ws4">s</div></div><div class="t m0 x22 h15 y64 ff4 fsc fc0 sc0 ls3a ws4">ns<span class="_ _11"></span>n<span class="_ _12"></span>w<span class="_ _13"></span>n<span class="_ _14"></span><span class="ff7 ls3b"> &#58888;</span></div><div class="t m0 xf h2 y65 ff1 fs0 fc0 sc0 ls22 ws66">where s(n) is speech signal, w(n) is sliding </div><div class="t m0 xe h2 y66 ff1 fs0 fc0 sc0 ls2b ws4">window. </div><div class="t m0 xf h2 y67 ff1 fs0 fc0 sc0 ls2a ws67">Short-time AMDF function-based pitch period </div><div class="t m0 xe h8 y68 ff1 fs0 fc0 sc0 ls2d ws4">initi<span class="ls22 ws68">al estimation is</span><span class="ff6 ls4">&#14822;</span></div><div class="t m4 x13 h16 y69 ff7 fsd fc0 sc0 ls3c ws4">&#58882;&#58883;</div><div class="t m0 x23 h17 y6a ff1 fse fc0 sc0 ls3d ws4">max</div><div class="t m0 x24 h17 y6b ff1 fse fc0 sc0 ls3e ws4">min</div><div class="t m0 x22 h18 y6c ff1 fsf fc0 sc0 ls3f ws4">arg</div><div class="t m0 x25 h19 y6d ff4 fs10 fc0 sc0 ls40 ws4">TP</div><div class="t m0 x26 h19 y6e ff4 fs10 fc0 sc0 ls41 ws4">kT<span class="_ _15"></span>P</div><div class="t m0 x27 h19 y6f ff4 fs10 fc0 sc0 ls4 ws4">k</div><div class="t m0 x28 h1a y70 ff4 fsf fc0 sc0 ls42 ws4">TP<span class="_ _16"> </span>D<span class="_ _17"> </span>k</div><div class="t m0 x26 h1a y71 ff4 fsf fc0 sc0 ls43 ws4">MIN</div><div class="t m0 x29 h1b y72 ff7 fs10 fc0 sc0 ls4 ws4"> </div><div class="t m0 x2a h1c y70 ff7 fsf fc0 sc0 ls4 ws4"> <span class="_ _18"> </span><span class="ff1 fs0 ls38"> <span class="_ _1"></span> (<span class="_ _d"></span>2<span class="_ _e"></span>)<span class="_ _d"></span> </span></div><div class="t m0 xf h2 y73 ff1 fs0 fc0 sc0 ls2a ws69">where TPmax and TPmin is pre=determined </div><div class="t m0 xe h2 y74 ff1 fs0 fc0 sc0 ls44 ws4">ma<span class="ls20 ws28">ximal and minimal pitch period, respectively. </span></div><div class="t m0 xf h2 y75 ff1 fs0 fc0 sc0 ls2a ws6a">When delay time increases, the peak ampli<span class="ls2b ws59">tude of </span></div><div class="t m0 xe h2 y76 ff1 fs0 fc0 sc0 ls1f ws26">short-time AMDF function <span class="ls2e ws6b">decreased gradually [3]. </span></div><div class="t m0 xe h2 y77 ff1 fs0 fc0 sc0 ls28 ws6c">This leads to the difficulty in detecting valley point </div><div class="t m0 xe h2 y78 ff1 fs0 fc0 sc0 ls23 ws3b">and apparent degree of valley points. And the </div><div class="t m0 xe h2 y79 ff1 fs0 fc0 sc0 ls1d ws6d">probability of double misjudge and half misjudge is </div><div class="t m0 xe h2 y7a ff1 fs0 fc0 sc0 ls1f ws27">very high. </div><div class="t m0 xf h2 y7b ff1 fs0 fc0 sc0 ls2a ws6e">LV-AMDF pitch detection algorithm<span class="_ _0"></span> is defined </div><div class="t m0 xe h8 y7c ff1 fs0 fc0 sc0 ls2d ws4">as<span class="ff6 ls4">&#14822;</span></div><div class="t m0 x2b h1d y7d ff1 fs11 fc0 sc0 ls4 ws4"> </div><div class="t m0 x1 h1e y7e ff1 fs12 fc0 sc0 ls45 ws4">_____________________________</div><div class="t m0 x1 h1f y7f ff2 fs13 fc0 sc0 ls46 ws6f">978-1-4244-3864-8/09/$25.00 &#169;2009 IEEE </div><div class="t m0 x2c h20 y80 ff5 fs14 fc0 sc0 ls4 ws4">&#58903;&#58896;&#58901;&#58907;&#58902;</div></div><div class="pi" data-data='{"ctm":[1.611639,0.000000,0.000000,1.611639,0.000000,0.000000]}'></div></div> </body> </html>
评论
    相关推荐
    • midi-and-mp3.rar
      面向midi音乐和mp3音乐的哼唱检索系统
    • DEMO.zip
      我们设计并实现了一种用于实时检测语音基频(pitch)、提取音乐旋律(melody)的软件。 语音或者音乐的基频是一种很重要的特征,可以用在很多研究与应用中。比如:用于普通话语音声调的识别和确认;语音流利度的...
    • codesforimageprocessing.rar
      实现简单图像处理,包括256色转灰度图、Hough变换、Walsh变换、中值滤波、二值化变换、亮度增减、傅立叶变换、反色、取对数、取指数、图像平移、图像旋转、图像细化、图像缩放、图像镜像、均值滤波、对比度拉伸、拉普拉斯锐化(边缘检测)、方块编码、梯度锐化、灰度均衡、直方图均衡、离散余弦变换、维纳滤波处理、逆滤波处理、阈值变换、高斯平滑。
    • matlab.rar
      基于MATLAB-GUI图形界面的数字图像处理软件 本系统设计基于GUI图形界面,用matlab语言编写代码,实现功能包括图象的读取、存储、显示、直方图均衡化、阈值化、小波分解、小波重构、加噪、去噪、平滑、锐化、边缘检测、图像分割等
    • OPENCV_SIFT_VC6.rar
      基于OPENCV的SIFT特征提取与匹配算法。包含完整的从图像高斯金字塔、DOG、空间极值点提取、关键点描述、KDtree匹配等关键步骤的全部函数实现,对全面深入理解Lowe的SIFT算法有莫大帮助。程序运行前须安装(1)OpenCV: http://opencvlibrary.sourceforge.net (2)SIFT: http://web.engr.oregonstate.edu/~hess/index.html,并配置其环境参数。
    • MATLAB_code_of_image_processing.rar
      该gui函数基本上包括图像处理里面的最基本处理,相当于一个小型photoshop。比如读取文件,几何变换中的垂直镜像,平移,旋转,缩放;正交变换的DFT,FFT,DCT,DST,DHT,DWashT;灰度处理中的反色,直方图均衡,全局线性变换,分段线性变换,指数非线性变换,对数非线性变换;图像增强里面的加噪声,平滑,锐化,伪彩色增强;图像分割里面的灰度阈值法,Robert,Laplace,sobel,prewitt,canny算子边缘检测法;图像恢复里面的直接逆滤波,维纳滤波;图像编码里面的霍夫曼编码,行程编码等等
    • bianyuan_jiance_bijiao.rar
      用六种算子(分别是gabor、拉普拉斯、priwitt、robert、sobel、wallis),对三种图象进行边缘检测比较,强烈推荐哦,是本人的毕业设计中的关键程序
    • 一维和二维小波变换的VC实现.rar
      小波算法在vc的具体实现.小波算法在图像处理,视频压缩,特征提取等领域有广泛应用,程序以灰度图像为数据分别演示了一维和二维小波变换,希望对大家有所帮助.谢谢!
    • thresold.rar
      基于小波变换阈值去噪的MATLAB源代码(包含小波包去噪程序)
    • HOG.rar
      基于HOG的行人检测,作者的源代码有错误,现代码已经进行了改变,并可以调试通