<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/625725b6bd8c6f2306b602d6/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/625725b6bd8c6f2306b602d6/bg1.jpg"><div class="t m0 x1 h2 y1 ff1 fs0 fc0 sc0 ls0 ws0">107</div><div class="t m0 x2 h3 y2 ff1 fs1 fc0 sc0 ls1 ws0">CHAPTER</div><div class="t m0 x3 h4 y3 ff1 fs2 fc0 sc0 ls2 ws0">6</div><div class="t m0 x4 h5 y4 ff2 fs3 fc0 sc0 ls3 ws1">Convolution </div><div class="t m0 x5 h6 y5 ff1 fs4 fc0 sc0 ls4 ws2">Convolution is a mathematical way of combining two signals to form a third signal. It is the</div><div class="t m0 x5 h6 y6 ff1 fs4 fc0 sc0 ls5 ws3">single most important technique in Digital Signal Processing. Using the strategy of impulse</div><div class="t m0 x5 h6 y7 ff1 fs4 fc0 sc0 ls6 ws4">decomposition, systems are described by a signal called the <span class="ff3 ls7 ws5">impulse response</span><span class="ls8 ws6">. Convolution is</span></div><div class="t m0 x5 h6 y8 ff1 fs4 fc0 sc0 ls9 ws7">important because it relates the three signals of interest: the input signal, the output signal, and</div><div class="t m0 x5 h6 y9 ff1 fs4 fc0 sc0 ls8 ws8">the impulse response. This chapter presents convolution from two different viewpoints, called</div><div class="t m0 x5 h6 ya ff1 fs4 fc0 sc0 lsa ws9">the input side algorithm and the output side algorithm. Convolution provides the mathematical</div><div class="t m0 x5 h6 yb ff1 fs4 fc0 sc0 lsb wsa">framework for DSP; there is nothing more important in this book. </div><div class="t m0 x5 h7 yc ff2 fs5 fc0 sc0 lsc wsb">The Delta Function and Impulse Response</div><div class="t m0 x6 h8 yd ff1 fs6 fc0 sc0 lsd wsc">The previous chapter describes how a signal can be decomposed into a group</div><div class="t m0 x6 h9 ye ff1 fs6 fc0 sc0 lse wsd">of components called <span class="ff4 lsf ws0">impulses</span><span class="ls10 wse">. An impulse is a signal composed of all zeros,</span></div><div class="t m0 x6 h8 yf ff1 fs6 fc0 sc0 ls11 wsf">except a single nonzero point. In effect, impulse decomposition provides a way</div><div class="t m0 x6 h8 y10 ff1 fs6 fc0 sc0 ls12 ws10">to analyze signals one sample at a time. The previous chapter also presented</div><div class="t m0 x6 h8 y11 ff1 fs6 fc0 sc0 ls13 ws11">the fundamental concept of DSP: the input signal is decomposed into simple</div><div class="t m0 x6 h8 y12 ff1 fs6 fc0 sc0 ls14 ws12">additive components, each of these components is passed through a linear</div><div class="t m0 x6 h8 y13 ff1 fs6 fc0 sc0 ls15 ws13">system, and the resulting output components are synthesized (added). The</div><div class="t m0 x6 h8 y14 ff1 fs6 fc0 sc0 ls16 ws14">signal resulting from this divide-and-conquer procedure is identical to that</div><div class="t m0 x6 h8 y15 ff1 fs6 fc0 sc0 ls17 ws15">obtained by directly passing the original signal through the system. While</div><div class="t m0 x6 h8 y16 ff1 fs6 fc0 sc0 ls18 ws16">many different decompositions are possible, two form the backbone of signal</div><div class="t m0 x6 h8 y17 ff1 fs6 fc0 sc0 ls19 ws17">processing: impulse decomposition and Fourier decomposition. When impulse</div><div class="t m0 x6 h8 y18 ff1 fs6 fc0 sc0 ls1a ws18">decomposition is used, the procedure can be described by a mathematical</div><div class="t m0 x6 h9 y19 ff1 fs6 fc0 sc0 ls1b ws19">operation called <span class="ff4 ls1c ws0">convolution</span><span class="ls1d ws1a">. In this chapter (and most of the following ones)</span></div><div class="t m0 x6 h8 y1a ff1 fs6 fc0 sc0 ls1e ws1b">we will only be dealing with <span class="ff3 ls1f ws0">discrete</span><span class="ls20 ws1c"> signals. Convolution also applies to</span></div><div class="t m0 x6 h8 y1b ff3 fs6 fc0 sc0 ls21 ws0">continuous<span class="ff1 ls22 ws1d"> signals, but the mathematics is more complicated. We will look at</span></div><div class="t m0 x6 h8 y1c ff1 fs6 fc0 sc0 ls23 ws1e">how continious signals are processed in Chapter 13. </div><div class="t m0 x6 h9 y1d ff1 fs6 fc0 sc0 ls24 ws1f">Figure 6-1 defines two important terms used in DSP. The first is the <span class="ff4 ls25 ws0">delta</span></div><div class="t m0 x6 h9 y1e ff4 fs6 fc0 sc0 ls26 ws0">function<span class="ff1 ls27 ws20">, symbolized by the Greek letter delta, <span class="_ _0"> </span><span class="ls28 ws21">. The delta function is</span></span></div><div class="c x7 y1f w2 ha"><div class="t m0 x8 hb y20 ff5 fs0 fc0 sc0 ls29 ws0">*<span class="_ _1"></span>*</div></div><div class="t m0 x9 hc y21 ff4 fs0 fc0 sc0 ls2a ws0">[<span class="ff6 ls2b">n</span>]</div><div class="t m0 x6 h8 y22 ff1 fs6 fc0 sc0 ls2c ws22">a <span class="ff3 ls2d ws0">normalized</span><span class="ls2e ws23"> impulse, that is, sample number zero has a value of one, while</span></div></div><div class="pi" data-data='{"ctm":[1.839080,0.000000,0.000000,1.839080,0.000000,0.000000]}'></div></div>
</body>
</html>
<div id="pf2" class="pf w0 h0" data-page-no="2"><div class="pc pc2 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/625725b6bd8c6f2306b602d6/bg2.jpg"><div class="t m0 xa h2 y23 ff3 fs0 fc0 sc0 ls2f ws24">The Scientist and Engineer's Guide to Digital Signal Processing<span class="_ _2"></span><span class="ff1 ls0 ws0">108</span></div><div class="t m0 xb h8 y24 ff1 fs6 fc0 sc0 ls30 ws25">all other samples have a value of zero. For this reason, the delta function is</div><div class="t m0 xb h9 y25 ff1 fs6 fc0 sc0 ls31 ws26">frequently called the <span class="ff4 ls32 ws27">unit impulse</span><span class="ls2d ws28">. </span></div><div class="t m0 xb h9 y26 ff1 fs6 fc0 sc0 ls33 ws29">The second term defined in Fig. 6-1 is the <span class="ff4 ls34 ws2a">impulse response</span><span class="ls35 ws2b">. As the name</span></div><div class="t m0 xb h8 y27 ff1 fs6 fc0 sc0 ls36 ws2c">suggests, the impulse response is the signal that exits a system when a delta</div><div class="t m0 xb h8 y28 ff1 fs6 fc0 sc0 ls37 ws2d">function (unit impulse) is the input. If two systems are different in any way,</div><div class="t m0 xb h8 y29 ff1 fs6 fc0 sc0 ls38 ws2e">they will have different impulse responses. Just as the input and output signals</div><div class="t m0 xb h8 y2a ff1 fs6 fc0 sc0 ls39 ws2f">are often called <span class="_ _3"> </span><span class="ls3a ws30"> and <span class="_ _3"> </span><span class="ls3b ws31">, the impulse response is usually given the<span class="_ _4"></span><span class="ff3 fs0 ls3c ws0">x<span class="ff1 ls2a">[</span><span class="ls0">n<span class="ff1 ls2a">]<span class="_ _5"> </span></span></span>y<span class="ff1 ls2a">[</span><span class="ls0">n<span class="ff1 ls2a">]</span></span></span></span></span></div><div class="t m0 xb h8 y2b ff1 fs6 fc0 sc0 ls3d ws32">symbol, <span class="_ _6"> </span><span class="ls3e ws33">. Of course, this can be changed if a more descriptive name is</span></div><div class="t m0 xc hc y2c ff6 fs0 fc0 sc0 ls2b ws0">h<span class="ff4 ls2a">[</span>n<span class="ff4 ls2a">]</span></div><div class="t m0 xb h8 y2d ff1 fs6 fc0 sc0 ls3f ws34">available, for instance, <span class="_ _0"> </span><span class="ls40 ws35"> might be used to identify the impulse response of<span class="_ _7"></span><span class="ff3 fs0 ls41 ws0">f<span class="_ _8"> </span><span class="ff1 ls2a">[</span><span class="ls0">n<span class="ff1 ls2a">]</span></span></span></span></div><div class="t m0 xb h8 y2e ff1 fs6 fc0 sc0 ls2c ws36">a <span class="ff3 ls42 ws0">filter<span class="ff1 ls2d">.</span></span></div><div class="t m0 xb h8 y2f ff1 fs6 fc0 sc0 ls43 ws37">Any impulse can be represented as a <span class="ff3 ls44 ws0">shifted</span><span class="ls3a ws30"> and <span class="ff3 ls45 ws0">scaled</span><span class="ls46 ws38"> delta function.</span></span></div><div class="t m0 xb h8 y30 ff1 fs6 fc0 sc0 ls47 ws39">Consider a signal, <span class="_ _0"> </span><span class="ls48 ws3a">, composed of all zeros except sample number 8,<span class="_ _9"></span><span class="ff3 fs0 ls0 ws0">a<span class="ff1 ls2a">[</span>n<span class="ff1 ls2a">]</span></span></span></div><div class="t m0 xb h8 y31 ff1 fs6 fc0 sc0 ls49 ws3b">which has a value of -3. This is the same as a delta function shifted to the</div><div class="t m0 xb h8 y32 ff1 fs6 fc0 sc0 ls4a ws3c">right by 8 samples, and multiplied by -3. In equation form:</div><div class="t m0 xd h8 y33 ff1 fs6 fc0 sc0 ls4b ws3d">. Make sure you understand this notation, it is used in<span class="_ _a"></span><span class="ff3 fs0 ls0 ws0">a<span class="ff1 ls2a">[</span>n<span class="ff1 ls2a">]<span class="_ _8"> </span><span class="ff7 ls4c">'<span class="_ _8"> </span>&</span><span class="ls0">3<span class="ff8 ls4d">*</span></span>[</span>n<span class="ff7 ls4c">&</span><span class="ff1">8<span class="ls2a">]</span></span></span></div><div class="t m0 xb h8 y34 ff1 fs6 fc0 sc0 ls4e ws3e">nearly all DSP equations. </div><div class="t m0 xb h8 y35 ff1 fs6 fc0 sc0 ls4f ws3f">If the input to a system is an impulse, such as <span class="_ _b"> </span><span class="ls50 ws40">, what is the system's<span class="_ _c"></span><span class="ff7 fs0 ls4c ws0">&<span class="ff1 ls0">3<span class="ff8 ls4d">*</span><span class="ls2a">[</span><span class="ff3">n</span></span>&<span class="ff1 ls0">8<span class="ls2a">]</span></span></span></span></div><div class="t m0 xb h8 y36 ff1 fs6 fc0 sc0 ls51 ws41">output? This is where the properties of homogeneity and shift invariance are</div><div class="t m0 xb h8 y37 ff1 fs6 fc0 sc0 ls52 ws42">used. Scaling and shifting the input results in an identical scaling and shifting</div><div class="t m0 xb h8 y38 ff1 fs6 fc0 sc0 ls53 ws43">of the output. If <span class="_ _3"> </span><span class="ls54 ws44"> results in <span class="_ _0"> </span><span class="ls55 ws45">, it follows that <span class="_ _b"> </span><span class="ls56 ws46"> results in<span class="_ _d"></span><span class="ff8 fs0 ls4d ws0">*<span class="ff1 ls2a">[<span class="ff3 ls0">n</span>]<span class="_ _e"> </span><span class="ff3 ls0">h</span>[<span class="ff3 ls0">n</span>]<span class="_ _f"> </span><span class="ff7 ls4c">&</span><span class="ls0">3</span></span>*<span class="ff1 ls2a">[<span class="ff3 ls0">n<span class="ff7 ls4c">&</span><span class="ff1">8</span></span>]</span></span></span></span></span></div><div class="t m0 xe h8 y39 ff1 fs6 fc0 sc0 ls57 ws47">. In words, the output is a version of the impulse response that has<span class="_ _a"></span><span class="ff7 fs0 ls4c ws0">&<span class="ff1 ls0">3<span class="ff3">h</span><span class="ls2a">[</span><span class="ff3">n</span></span>&<span class="ff1 ls0">8<span class="ls2a">]</span></span></span></div><div class="t m0 xb h8 y3a ff1 fs6 fc0 sc0 ls58 ws48">been <span class="ff3 ls59 ws49">shifted </span><span class="ls5a ws4a">and <span class="ff3 ls5b ws0">scaled</span><span class="ls5c ws4b"> by the same amount as the delta function on the input.</span></span></div><div class="t m0 xb h8 y3b ff1 fs6 fc0 sc0 ls5d ws4c">If you know a system's impulse response, you immediately know how it will</div><div class="t m0 xb h8 y3c ff1 fs6 fc0 sc0 ls5e ws4d">react to <span class="ff3 ls5f ws0">any</span><span class="ls60 ws4e"> impulse.</span></div><div class="t m0 xf h7 y3d ff2 fs5 fc0 sc0 ls61 ws0">Convolution</div><div class="t m0 xb h8 y3e ff1 fs6 fc0 sc0 ls62 ws4f">Let's summarize this way of understanding how a system changes an input</div><div class="t m0 xb h8 y3f ff1 fs6 fc0 sc0 ls63 ws50">signal into an output signal. First, the input signal can be decomposed into a</div><div class="t m0 xb h8 y40 ff1 fs6 fc0 sc0 ls64 ws51">set of impulses, each of which can be viewed as a scaled and shifted delta</div><div class="t m0 xb h8 y41 ff1 fs6 fc0 sc0 ls65 ws52">function. Second, the output resulting from each impulse is a scaled and shifted</div><div class="t m0 xb h8 y42 ff1 fs6 fc0 sc0 ls52 ws42">version of the impulse response. Third, the overall output signal can be found</div><div class="t m0 xb h8 y43 ff1 fs6 fc0 sc0 ls66 ws53">by adding these scaled and shifted impulse responses. In other words, if we</div><div class="t m0 xb h8 y44 ff1 fs6 fc0 sc0 ls67 ws54">know a system's impulse response, then we can calculate what the output will</div><div class="t m0 xb h8 y45 ff1 fs6 fc0 sc0 ls68 ws55">be for any possible input signal. This means we know <span class="ff3 ls69 ws0">everything</span><span class="ls6a ws56"> about the</span></div><div class="t m0 xb h8 y46 ff1 fs6 fc0 sc0 ls6b ws57">system. There is nothing more that can be learned about a linear system's</div><div class="t m0 xb h8 y47 ff1 fs6 fc0 sc0 ls6c ws58">characteristics. (However, in later chapters we will show that this information</div><div class="t m0 xb h8 y48 ff1 fs6 fc0 sc0 ls6d ws59">can be represented in different forms). </div><div class="t m0 xb h8 y49 ff1 fs6 fc0 sc0 ls6e ws5a">The impulse response goes by a different name in some applications. If the</div><div class="t m0 xb h9 y4a ff1 fs6 fc0 sc0 ls6f ws5b">system being considered is a <span class="ff3 ls42 ws0">filter</span><span class="ls70 ws5c">, the impulse response is called the <span class="ff4 ls71 ws0">filter</span></span></div><div class="t m0 xb h9 y4b ff4 fs6 fc0 sc0 ls72 ws0">kernel<span class="ff1 ls6a ws5d">, the </span><span class="ls73 ws5e">convolution kernel<span class="ff1 ls74 ws5f">, or simply, the </span></span>kernel<span class="ff1 ls75 ws60">. In image processing,</span></div><div class="t m0 xb h9 y4c ff1 fs6 fc0 sc0 ls76 ws61">the impulse response is called the <span class="ff4 ls77 ws62">point spread function</span><span class="ls78 ws63">. While these terms</span></div><div class="t m0 xb h8 y4d ff1 fs6 fc0 sc0 ls56 ws64">are used in slightly different ways, they all mean the same thing, the signal</div><div class="t m0 xb h8 y4e ff1 fs6 fc0 sc0 ls79 ws65">produced by a system when the input is a delta function.</div></div><div class="pi" data-data='{"ctm":[1.839080,0.000000,0.000000,1.839080,0.000000,0.000000]}'></div></div>
<div id="pf3" class="pf w0 h0" data-page-no="3"><div class="pc pc3 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/625725b6bd8c6f2306b602d6/bg3.jpg"><div class="t m0 x10 h2 y23 ff3 fs0 fc0 sc0 ls7a ws66">Chapter 6- Convolution<span class="_ _10"> </span><span class="ff1 ls0 ws0">109</span></div><div class="t m0 x11 hd y4f ff9 fs7 fc0 sc0 ls7b ws0">System</div><div class="t m0 x12 he y50 ff1 fs8 fc0 sc0 ls7c ws0">-2<span class="_"> </span>-1<span class="_ _11"> </span><span class="ls7d">0<span class="_ _12"> </span>1<span class="_ _12"> </span>2<span class="_ _12"> </span>3<span class="_ _12"> </span>4<span class="_ _12"> </span>5<span class="_ _12"> </span>6</span></div><div class="t m0 xa he y51 ff1 fs8 fc0 sc0 ls7c ws0">-1</div><div class="t m0 x13 he y52 ff1 fs8 fc0 sc0 ls7d ws0">0</div><div class="t m0 x13 he y4 ff1 fs8 fc0 sc0 ls7d ws0">1</div><div class="t m0 x13 he y53 ff1 fs8 fc0 sc0 ls7d ws0">2</div><div class="t m0 x14 hf y54 ff1 fs9 fc0 sc0 ls7e ws0">-2<span class="_ _8"> </span>-1<span class="_ _11"> </span><span class="ls0">0<span class="_ _12"> </span>1<span class="_ _12"> </span>2<span class="_ _13"> </span>3<span class="_ _12"> </span>4<span class="_ _13"> </span>5<span class="_ _12"> </span>6</span></div><div class="t m0 x15 hf y55 ff1 fs9 fc0 sc0 ls7e ws0">-1</div><div class="t m0 x16 hf y56 ff1 fs9 fc0 sc0 ls0 ws0">0</div><div class="t m0 x16 hf y4 ff1 fs9 fc0 sc0 ls0 ws0">1</div><div class="t m0 x16 hf y57 ff1 fs9 fc0 sc0 ls0 ws0">2</div><div class="t m0 x17 h10 y58 ff8 fs7 fc0 sc0 ls7f ws0">*<span class="ff9 ls80">[n]</span></div><div class="t m0 x18 hd y59 ff9 fs7 fc0 sc0 ls81 ws0">h[n]</div><div class="t m0 x19 hd y5a ff9 fs7 fc0 sc0 ls82 ws0">Delta</div><div class="t m0 x1a hd y5b ff9 fs7 fc0 sc0 ls83 ws0">Impulse</div><div class="t m0 x1b hd y5c ff9 fs7 fc0 sc0 ls84 ws0">Response</div><div class="t m0 x1c hd y5d ff9 fs7 fc0 sc0 ls85 ws0">Linear</div><div class="t m0 x1d hd y5c ff9 fs7 fc0 sc0 ls86 ws0">Function</div><div class="t m0 x1e h11 y5e ff1 fsa fc0 sc0 ls87 ws67">FIGURE 6-1</div><div class="t m0 x1e h11 y5f ff1 fsa fc0 sc0 ls88 ws68">Definition of <span class="ff3 ls89 ws69">delta function</span><span class="ls7d ws6a"> <span class="ls8a ws6b">and <span class="ff3 ls8b ws6c">impulse response</span><span class="ls8c ws6d">. The delta function is a normalized impulse. All of</span></span></span></div><div class="t m0 x1e h11 y60 ff1 fsa fc0 sc0 ls8d ws6e">its samples have a value of zero, except for sample number zero, which has a value of one. The Greek</div><div class="t m0 x1e h11 y61 ff1 fsa fc0 sc0 ls58 ws6f">letter delta, <span class="_ _14"> </span><span class="ls8e ws70">, is used to identify the delta function. The <span class="ff3 ls8b ws71">impulse response</span><span class="ls8f ws72"> of a linear system, usually</span></span></div><div class="t m0 x1f h12 y62 ff8 fsb fc0 sc0 ls90 ws0">*<span class="ff1 ls91">[<span class="ff3 ls7d">n</span>]</span></div><div class="t m0 x1e h11 y63 ff1 fsa fc0 sc0 ls92 ws73">denoted by <span class="_ _15"> </span><span class="ls93 ws74">, is the output of the system when the input is a delta function.<span class="_ _16"></span><span class="ff3 fsb ls7d ws0">h<span class="ff1 ls91">[</span>n<span class="ff1 ls91">]</span></span></span></div><div class="t m0 x20 h13 y64 ff1 fsc fc0 sc0 ls94 ws75">x[n] h[n] = y[n]</div><div class="t m0 x21 h13 y65 ff1 fsc fc0 sc0 ls95 ws0">x[n]</div><div class="t m0 x22 h13 y66 ff1 fsc fc0 sc0 ls95 ws0">y[n]</div><div class="t m0 x23 h13 y67 ff1 fsc fc0 sc0 ls96 ws0">Linear</div><div class="t m0 x24 h13 y68 ff1 fsc fc0 sc0 ls97 ws0">System</div><div class="t m0 x16 h13 y69 ff1 fsc fc0 sc0 ls95 ws0">h[n]</div><div class="t m0 x5 h11 y6a ff1 fsa fc0 sc0 ls87 ws67">FIGURE 6-2</div><div class="t m0 x5 h11 y6b ff1 fsa fc0 sc0 ls98 ws76">How convolution is used in DSP. The</div><div class="t m0 x5 h11 y6c ff1 fsa fc0 sc0 ls99 ws77">output signal from a linear system is</div><div class="t m0 x5 h11 y4a ff1 fsa fc0 sc0 ls9a ws78">equal to the input signal <span class="ff3 ls9b ws0">convolved</span></div><div class="t m0 x5 h11 y6d ff1 fsa fc0 sc0 ls9c ws79">with the system's impulse response.</div><div class="t m0 x5 h11 y6e ff1 fsa fc0 sc0 ls9d ws7a">Convolution is denoted by a star when</div><div class="t m0 x5 h11 y6f ff1 fsa fc0 sc0 ls9e ws7b">writing equations. </div><div class="t m0 x6 h8 y70 ff1 fs6 fc0 sc0 ls9f ws7c">Convolution is a formal mathematical operation, just as multiplication,</div><div class="t m0 x6 h8 y71 ff1 fs6 fc0 sc0 lsa0 ws7d">addition, and integration. Addition takes two <span class="ff3 lsa1 ws0">numbers</span><span class="lsa2 ws7e"> and produces a third</span></div><div class="t m0 x6 h8 y72 ff3 fs6 fc0 sc0 ls5b ws0">number<span class="ff1 lsa3 ws7f">, while convolution takes two </span><span class="lsa4">signals<span class="ff1 lsa5 ws80"> and produces a third </span><span class="lsa6">signal<span class="ff1 ls2d">.</span></span></span></div><div class="t m0 x6 h8 y73 ff1 fs6 fc0 sc0 lsa7 ws81">Convolution is used in the mathematics of many fields, such as probability and</div><div class="t m0 x6 h8 y74 ff1 fs6 fc0 sc0 lsa8 ws82">statistics. In linear systems, convolution is used to describe the relationship</div><div class="t m0 x6 h8 y75 ff1 fs6 fc0 sc0 lsa9 ws83">between three signals of interest: the input signal, the impulse response, and the</div><div class="t m0 x6 h8 y76 ff1 fs6 fc0 sc0 lsaa ws84">output signal.</div><div class="t m0 x6 h8 y77 ff1 fs6 fc0 sc0 lsab ws33">Figure 6-2 shows the notation when convolution is used with linear systems.</div><div class="t m0 x6 h8 y78 ff1 fs6 fc0 sc0 lsac ws85">An input signal, <span class="_ _3"> </span><span class="lsad ws86">, enters a linear system with an impulse response, <span class="_ _0"> </span><span class="ls2d ws0">,<span class="_ _17"></span><span class="ff3 fs0 ls3c">x<span class="ff1 ls2a">[</span><span class="ls0">n<span class="ff1 ls2a">]<span class="_ _18"> </span></span>h<span class="ff1 ls2a">[</span>n<span class="ff1 ls2a">]</span></span></span></span></span></div><div class="t m0 x6 h8 y79 ff1 fs6 fc0 sc0 lsae ws87">resulting in an output signal, <span class="_ _3"> </span><span class="lsaf ws88">. In equation form: <span class="_ _f"> </span><span class="ls2d ws0">.<span class="_ _19"></span><span class="ff3 fs0 ls3c">y<span class="ff1 ls2a">[</span><span class="ls0">n<span class="ff1 ls2a">]<span class="_ _1a"> </span></span></span>x<span class="ff1 ls2a">[</span><span class="ls0">n<span class="ff1 ls2a">]<span class="_ _8"> </span><span class="ffa lsb0">t<span class="_ _8"> </span></span></span>h<span class="ff1 ls2a">[</span>n<span class="ff1 ls2a">]<span class="_"> </span><span class="ff7 ls4c">'<span class="_ _8"> </span></span></span></span>y<span class="ff1 ls2a">[</span><span class="ls0">n<span class="ff1 ls2a">]</span></span></span></span></span></div><div class="t m0 x6 h8 y7a ff1 fs6 fc0 sc0 lsb1 ws89">Expressed in words, the input signal convolved with the impulse response is</div><div class="t m0 x6 h8 y7b ff1 fs6 fc0 sc0 lsb2 ws8a">equal to the output signal. Just as addition is represented by the plus, +, and</div><div class="t m0 x6 h8 y7c ff1 fs6 fc0 sc0 lsb3 ws8b">multiplication by the cross, ×, convolution is represented by the star, <span class="ffa fs4 lsb4 ws0">t</span><span class="ls5e ws8c">. It is</span></div><div class="t m0 x6 h8 y7d ff1 fs6 fc0 sc0 lsb5 ws8d">unfortunate that most programming languages also use the star to indicate</div><div class="t m0 x6 h8 y7e ff1 fs6 fc0 sc0 lsb6 ws8e">multiplication. A star in a computer program means multiplication, while a star</div><div class="t m0 x6 h8 y7f ff1 fs6 fc0 sc0 lsb7 ws8f">in an equation means convolution. </div></div><div class="pi" data-data='{"ctm":[1.839080,0.000000,0.000000,1.839080,0.000000,0.000000]}'></div></div>
<div id="pf4" class="pf w0 h0" data-page-no="4"><div class="pc pc4 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/625725b6bd8c6f2306b602d6/bg4.jpg"><div class="t m0 xa h2 y23 ff3 fs0 fc0 sc0 ls2f ws24">The Scientist and Engineer's Guide to Digital Signal Processing<span class="_ _2"></span><span class="ff1 ls0 ws0">110</span></div><div class="t m0 x25 h14 y80 ff9 fs8 fc0 sc0 lsb8 ws90">Sample number</div><div class="t m0 x26 h15 y81 ff1 fsd fc0 sc0 ls7d ws0">0<span class="_ _1b"> </span>10<span class="_ _1c"> </span>20<span class="_ _1c"> </span>30<span class="_ _1c"> </span>40<span class="_ _1c"> </span>50<span class="_ _1c"> </span>60<span class="_ _1c"> </span>70<span class="_ _1c"> </span>80<span class="_ _1c"> </span>90<span class="_ _1d"> </span>100<span class="_ _13"> </span>110</div><div class="t m0 x27 h15 y82 ff1 fsd fc0 sc0 lsb9 ws0">-2</div><div class="t m0 x27 h15 y83 ff1 fsd fc0 sc0 lsb9 ws0">-1</div><div class="t m0 x28 h15 y84 ff1 fsd fc0 sc0 ls7d ws0">0</div><div class="t m0 x28 h15 y85 ff1 fsd fc0 sc0 ls7d ws0">1</div><div class="t m0 x28 h15 y86 ff1 fsd fc0 sc0 ls7d ws0">2</div><div class="t m0 x28 h15 y87 ff1 fsd fc0 sc0 ls7d ws0">3</div><div class="t m0 x28 h15 y88 ff1 fsd fc0 sc0 ls7d ws0">4</div><div class="t m0 x29 h14 y89 ff9 fs8 fc0 sc0 lsba ws0"><span class="fc1 sc0">S</span></div><div class="t m0 x2a h15 y8a ff1 fsd fc0 sc0 ls7d ws0">0<span class="_ _1c"> </span>10<span class="_ _1d"> </span>20<span class="_ _1e"> </span>30</div><div class="t m0 x2b h15 ya ff1 fsd fc0 sc0 lsbb ws0">-0.25</div><div class="t m0 x2c h15 y8b ff1 fsd fc0 sc0 lsbc ws0">0.00</div><div class="t m0 x2c h15 y8c ff1 fsd fc0 sc0 lsbc ws0">0.25</div><div class="t m0 x2c h15 y85 ff1 fsd fc0 sc0 lsbc ws0">0.50</div><div class="t m0 x2c h15 y86 ff1 fsd fc0 sc0 lsbc ws0">0.75</div><div class="t m0 x2c h15 y87 ff1 fsd fc0 sc0 lsbc ws0">1.00</div><div class="t m0 x2c h15 y88 ff1 fsd fc0 sc0 lsbc ws0">1.25</div><div class="t m0 x29 h14 y8d ff9 fs8 fc0 sc0 lsba ws0">S</div><div class="t m0 x2a h15 y8e ff1 fsd fc0 sc0 ls7d ws0">0<span class="_ _1c"> </span>10<span class="_ _1d"> </span>20<span class="_ _1d"> </span>30</div><div class="t m0 x2b h15 y8f ff1 fsd fc0 sc0 lsbb ws0">-0.02</div><div class="t m0 x2c h15 y90 ff1 fsd fc0 sc0 lsbc ws0">0.00</div><div class="t m0 x2c h15 y91 ff1 fsd fc0 sc0 lsbc ws0">0.02</div><div class="t m0 x2c h15 y92 ff1 fsd fc0 sc0 lsbc ws0">0.04</div><div class="t m0 x2c h15 y93 ff1 fsd fc0 sc0 lsbc ws0">0.06</div><div class="t m0 x2c h15 y94 ff1 fsd fc0 sc0 lsbc ws0">0.08</div><div class="t m0 x2d h16 y95 ff9 fse fc0 sc0 lsbd ws91">a. Low-pass Filter</div><div class="t m0 x2d h16 y96 ff9 fse fc0 sc0 lsbe ws92">b. High-pass Filter</div><div class="t m0 x2e h14 y8d ff9 fs8 fc0 sc0 lsb8 ws90">Sample number</div><div class="t m0 x2 h15 y97 ff1 fsd fc0 sc0 ls7d ws0">0<span class="_ _1f"> </span>10<span class="_ _1e"> </span>20<span class="_ _20"> </span>30<span class="_ _1e"> </span>40<span class="_ _20"> </span>50<span class="_ _1e"> </span>60<span class="_ _20"> </span>70<span class="_ _1e"> </span>80</div><div class="t m0 x2f h15 y8f ff1 fsd fc0 sc0 lsb9 ws0">-2</div><div class="t m0 x2f h15 y98 ff1 fsd fc0 sc0 lsb9 ws0">-1</div><div class="t m0 x30 h15 y56 ff1 fsd fc0 sc0 ls7d ws0">0</div><div class="t m0 x30 h15 y99 ff1 fsd fc0 sc0 ls7d ws0">1</div><div class="t m0 x30 h15 y9a ff1 fsd fc0 sc0 ls7d ws0">2</div><div class="t m0 x30 h15 y9b ff1 fsd fc0 sc0 ls7d ws0">3</div><div class="t m0 x30 h15 y9c ff1 fsd fc0 sc0 ls7d ws0">4</div><div class="t m0 x25 h14 y9d ff9 fs8 fc0 sc0 lsbf ws93">Sample number</div><div class="t m0 x26 h15 y97 ff1 fsd fc0 sc0 ls7d ws0">0<span class="_ _1b"> </span>10<span class="_ _1c"> </span>20<span class="_ _1c"> </span>30<span class="_ _1c"> </span>40<span class="_ _1c"> </span>50<span class="_ _1c"> </span>60<span class="_ _1c"> </span>70<span class="_ _1c"> </span>80<span class="_ _1c"> </span>90<span class="_ _1e"> </span>100<span class="_ _21"> </span>110</div><div class="t m0 x27 h15 y9e ff1 fsd fc0 sc0 lsb9 ws0">-2</div><div class="t m0 x27 h15 y98 ff1 fsd fc0 sc0 lsb9 ws0">-1</div><div class="t m0 x28 h15 y56 ff1 fsd fc0 sc0 ls7d ws0">0</div><div class="t m0 x28 h15 y99 ff1 fsd fc0 sc0 ls7d ws0">1</div><div class="t m0 x28 h15 y9a ff1 fsd fc0 sc0 ls7d ws0">2</div><div class="t m0 x28 h15 y9b ff1 fsd fc0 sc0 ls7d ws0">3</div><div class="t m0 x28 h15 y9f ff1 fsd fc0 sc0 ls7d ws0">4</div><div class="t m0 x2e h14 y80 ff9 fs8 fc0 sc0 lsb8 ws90">Sample number</div><div class="t m0 x31 h15 y81 ff1 fsd fc0 sc0 ls7d ws0">0<span class="_ _1f"> </span>10<span class="_ _1e"> </span>20<span class="_ _1e"> </span>30<span class="_ _20"> </span>40<span class="_ _1e"> </span>50<span class="_ _1e"> </span>60<span class="_ _20"> </span>70<span class="_ _1e"> </span>80</div><div class="t m0 x2f h15 ya0 ff1 fsd fc0 sc0 lsb9 ws0">-2</div><div class="t m0 x2f h15 y83 ff1 fsd fc0 sc0 lsb9 ws0">-1</div><div class="t m0 x30 h15 y8c ff1 fsd fc0 sc0 ls7d ws0">0</div><div class="t m0 x30 h15 y85 ff1 fsd fc0 sc0 ls7d ws0">1</div><div class="t m0 x30 h15 y86 ff1 fsd fc0 sc0 ls7d ws0">2</div><div class="t m0 x30 h15 ya1 ff1 fsd fc0 sc0 ls7d ws0">3</div><div class="t m0 x30 h15 y88 ff1 fsd fc0 sc0 ls7d ws0">4</div><div class="t m0 x32 h17 y9d ff9 fsf fc0 sc0 lsc0 ws94">Sample number</div><div class="t m0 x32 h17 ya2 ff9 fsf fc0 sc0 lsc0 ws94">Sample number</div><div class="t m0 x2e h18 ya3 ff9 fs10 fc0 sc0 lsc1 ws95">Input Signal<span class="_ _22"> </span><span class="lsc2 ws96">Impulse Response<span class="_ _23"> </span><span class="lsc3 ws97">Output Signal</span></span></div><div class="t m1 x33 h19 ya4 ff1 fs11 fc0 sc0 ls7d ws0">Amplitude<span class="_ _24"></span>Amplitude</div><div class="t m1 x34 h19 ya4 ff1 fs11 fc0 sc0 ls7d ws0">Amplitude<span class="_ _24"></span>Amplitude</div><div class="t m1 x35 h19 ya4 ff1 fs11 fc0 sc0 ls7d ws0">Amplitude<span class="_ _24"></span>Amplitude</div><div class="t m0 x36 h11 ya5 ff1 fsa fc0 sc0 ls87 ws67">FIGURE 6-3</div><div class="t m0 x36 h11 ya6 ff1 fsa fc0 sc0 lsc4 ws98">Examples of low-pass and high-pass filtering using convolution. In this example, the input signal</div><div class="t m0 x36 h11 ya7 ff1 fsa fc0 sc0 lsc5 ws99">is a few cycles of a sine wave plus a slowly rising ramp. These two components are separated by</div><div class="t m0 x36 h11 ya8 ff1 fsa fc0 sc0 lsc6 ws9a">using properly selected impulse responses.</div><div class="t m0 xb h8 y3d ff1 fs6 fc0 sc0 lsc7 ws9b">Figure 6-3 shows convolution being used for low-pass and high-pass filtering.</div><div class="t m0 xb h8 ya9 ff1 fs6 fc0 sc0 lsc8 ws9c">The example input signal is the sum of two components: three cycles of a sine</div><div class="t m0 xb h8 y3e ff1 fs6 fc0 sc0 lsc9 ws9d">wave (representing a high frequency), plus a slowly rising ramp (composed of</div><div class="t m0 xb h8 y3f ff1 fs6 fc0 sc0 lsca ws9e">low frequencies). In (a), the impulse response for the low-pass filter is a</div><div class="t m0 xb h8 y40 ff1 fs6 fc0 sc0 lscb ws9f">smooth arch, resulting in only the slowly changing ramp waveform being</div><div class="t m0 xb h8 y41 ff1 fs6 fc0 sc0 lscc wsa0">passed to the output. Similarly, the high-pass filter, (b), allows only the more</div><div class="t m0 xb h8 y42 ff1 fs6 fc0 sc0 lscd wsa1">rapidly changing sinusoid to pass. </div><div class="t m0 xb h8 y44 ff1 fs6 fc0 sc0 lsce wsa2">Figure 6-4 illustrates two additional examples of how convolution is used to</div><div class="t m0 xb h8 y45 ff1 fs6 fc0 sc0 lscf wsa3">process signals. The inverting attenuator, (a), flips the signal top-for-bottom,</div><div class="t m0 xb h8 y46 ff1 fs6 fc0 sc0 lsd0 wsa4">and reduces its amplitude. The discrete derivative (also called the first</div><div class="t m0 xb h8 y47 ff1 fs6 fc0 sc0 lsd1 wsa5">difference), shown in (b), results in an output signal related to the <span class="ff3 lsd2 ws0">slope</span><span class="lsd3 wsa6"> of the</span></div><div class="t m0 xb h8 y48 ff1 fs6 fc0 sc0 lsd4 wsa7">input signal.</div><div class="t m0 xb h8 y49 ff1 fs6 fc0 sc0 lsd5 wsa8">Notice the lengths of the signals in Figs. 6-3 and 6-4. The input signals are</div><div class="t m0 xb h8 y4a ff1 fs6 fc0 sc0 lsd6 wsa9">81 samples long, while each impulse response is composed of 31 samples.</div><div class="t m0 xb h8 y4b ff1 fs6 fc0 sc0 ls36 wsaa">In most DSP applications, the input signal is hundreds, thousands, or even</div><div class="t m0 xb h8 y4c ff1 fs6 fc0 sc0 lsd7 wsab">millions of samples in length. The impulse response is usually much shorter,</div><div class="t m0 xb h8 y4d ff1 fs6 fc0 sc0 lsd8 wsac">say, a few points to a few hundred points. The mathematics behind</div><div class="t m0 xb h8 y4e ff1 fs6 fc0 sc0 lsd9 wsad">convolution doesn't restrict how long these signals are. It does, however,</div><div class="t m0 xb h8 yaa ff1 fs6 fc0 sc0 lsda wsae">specify the length of the output signal. The length of the output signal is</div></div><div class="pi" data-data='{"ctm":[1.839080,0.000000,0.000000,1.839080,0.000000,0.000000]}'></div></div>