Fourier Transformation f(x) Fourier Transformasjon F(u) 1 Continuous Fourier Transform Def The Fourier transform of a one-dimentional function f(x) fˆ (u ) f ( x)e j 2ux dx The Inverse Fourier Transform f ( x) fˆ (u )e j 2ux du 2 Continuous Fourier Transform Def - Notation The Fourier transform of a one-dimentional function f(x) fˆ (u ) F (u ) e j 2ux f ( x) F f ( x) e j 2ux f ( x)dx The inverse Fourier Transform of F(u) f ( x) e j 2ux fˆ (u ) F 1 fˆ (u ) e j 2ux fˆ (u )du 3 Continuous Fourier Transform Alternative Def F (u ) F f ( x) f ( x )e j 2ux dx F ( ) F f ( x) f ( x) F F (u) fˆ (u )e f ( x)e jx dx 1 j 2ux du 1 f ( x) F F ( ) 2 1 1 F ( ) F f ( x) 2 1 f ( x) F F ( ) 2 1 fˆ (u)e jx d f ( x)e jx dx fˆ (u )e jx d 4 Continuous Fourier Transform Example - cos(2ft) 5 Continuous Fourier Transform Example - cos(t) 6 Continuous Fourier Transform Example - sin(t) 7 Continuous Fourier Transform Example - Delta-function 8 Continuous Fourier Transform Example - Gauss function 9 Signals and Fourier Transform Frequency Information y1 sin( 1t ) FT FT y2 sin( 2t ) FT y3 sin( 1t ) sin( 2t ) 10 Stationary / Non-stationary signals Stationary FT y3 sin( 1t ) sin( 2t ) Non stationary sin( 1t ) y4 sin( 2t ) hvis t 60 FT hvis t 60 The stationary and the non-stationary signal both have the same FT. FT is not suitable to take care of non-stationary signals to give information about time. 11 Transient Signal Frequency Information Constant function in [-3,3]. Dominating frequency = 0 and some freequency because of edges. Transient signal resulting in extra frequencies > 0. Narrower transient signal resulting in extra higher frequencies pushed away from origin. 12 Transient Signal No Information about Position Moving the transient part of the signal to a new position does not result in any change in the transformed signal. Conclusion: The Fourier transformation contains information of a transient part of a signal, but only the frequency not the position. 13 fˆ (u ) f ( x )e j 2ux dx Inverse Fourier Transform [1/3] f ( x) fˆ (u)e Theorem: e t 2 e e jt dt Proof: f ( y ) e t e yt dt e t 2 y2 4 2 2 4 yt 0 dt e ( t y 2 y2 ) 2 4 dt 4y t e e dt e 1 2 2 y j e 4 jt e dt e 2 t 2 14 j 2ux du fˆ (u ) f ( x )e j 2ux dx Inverse Fourier Transform [2/3] f ( x) fˆ (u)e Theorem: f(y)gˆ(ay)dy - Proof: fˆ(ay)g(y)dy f , g L1 ( R) - j 2ayx ˆ f(y) g (ay)dy f(y) g ( x ) e dx dy - - f ( y ) g ( x )e j 2ayx dxdy f ( y ) g ( x)e j 2ayx dydx f ( y )e j 2ayx dy g ( x)dx fˆ (ax) g ( x)dx fˆ (ay) g ( y)dy 15 j 2ux du fˆ (u ) f ( x )e j 2ux dx Inverse Fourier Transform [3/3] f ( x) fˆ (u)e j 2ux du g ( x ) 1 2 gˆ (u ) e ( 2u ) e x2 4 f ( x) f ( x 0) 2 lim g ( x)dx 1 0 f ( y) g ( x y)dy lim f ( x) * g ( x) 0 lim 0 y ˆ f ( y ) g dy 2 lim 0 1 jtx t 2 g (t ) e e 2 1 jtx t 2 j 2yt gˆ ( y ) e e e dt 2 1 2 1 2 t j ( 2y x ) t dt e e 2 ( 24y x ) e g ( x 2y ) 2 ˆ y g ( y )dy f 2 1 lim 0 2 1 2 ˆ y e jyx e y 2 dy f 2 ˆ y f 2 jyx e dy 1 2 fˆ ( y )e j 2yx dy 2 F [ fˆ ( x)] 1 16 Properties F[ f g ] F[ f ] F[ g ] F [cf ] cF [ f ] F 1[ f g ] F 1[ f ] F 1[ g ] F 1[cf ] cF 1[ f ] F[ f (n) ] dn F [ f ] j n d n n d ( j ) F 1[ f ] n dt ( j ) n F [ f ] F 1[ f ( n ) ] ( jt ) n F 1[ f ] F [ f (t a )] e ja F [ f (t )] n F [t f ] F 1[ n f ] n F [ f (at )]( ) F [ f ]( ) 1 F [ f (t )] a a L[ f ]( j ) L[ f ] f (t )e ts dt 0 17 Fourier Transforms of Harmonic and Constant Function f ( x) F (u u0 ) (u u0 ) 1 j 2ux du e ) u u ( ) u u ( 0 0 (u u0 )e j 2ux du (u u0 )e j 2ux du 2 cos( 2u0 x) e j 2ux e j 2ux 2 j sin( 2u0 x) - 1 (u 0 ) (u u0 ) 2 j F sin( 2u0 x) (u u0 ) (u u0 ) 2 F cos( 2u0 x) F 1 (u ) 18 Fourier Transforms of Some Common Functions f ( x) e j 2u0 x F (u ) 1 (u u0 ) (u u0 ) 2 j δ(u u0 ) δ(u u0 ) 2 (u u0 ) δ(x) 1 cos(2u 0 x) sin (2u 0 x) ( x) ( x) sin 2 (u ) u sin 2 (u ) (u ) 2 u ( x) 1 j ( u ) 2 u e x e u 2 2 19 Even and Odd Functions [1/3] Def f even (t ) f even (t ) f odd (t ) f odd (t ) Every function can be split in an even and an odd part f (t ) f even (t ) f odd (t ) 1 f (t ) f (t ) 2 1 f (t ) f (t ) f odd (t ) 2 f even (t ) Every function can be split in an even and an odd part and each of this can in turn be split in a real and an imaginary part f (t ) f even (t ) f odd (t ) f even ,real (t ) f odd ,real (t ) f even ,imag (t ) f odd ,imag (t ) 20 Even and Odd Functions [2/3] F (u ) f ( x )e j 2ux dx f ( x) cos(2ux)dx j f (t ) sin( 2ux)dx f even ( x) cos( 2ux)dx f f odd ( x) cos( 2ux)dx j f even ( x) sin( 2ux)dx j f odd ( x) sin( 2ux)dx even ( x) cos( 2ux)dx j f odd ( x) sin( 2ux)dx Feven (u ) jFodd (u ) 1. 2. 3. Even component in f produces an even component in F Odd component in f produces an odd component in F Odd component in f produces an coefficient -j 21 Even and Odd Functions [3/3] f(t) F(u ) Even Odd Real Even Even Odd Real Even Real Odd Imag Odd Imag Even Imag Even Complex Even Complex Even Complex Odd Complex Odd Real Hermite Real Even plus Imag Odd Real Odd plus Imag Even Real Imag Hermite F (u ) F * (u ) 22 The Shift Theorem F f ( x a ) f ( x a )e j 2ux dx f ( x)e j 2u ( x a ) dx e j 2ua f ( x)e j 2ux dx F f ( x a) e j 2ua F f ( x) e j 2ua F (u ) e j 2ua F f ( x) e j 2ua F (u ) 23 The Similarity Theorem F f (ax) f (ax)e j 2ux dx 1 a f ( x )e 1 u F a a j 2 au x dx 1 u F f (ax) F a a 24 The Convolution Theorem f (t ) * g (t ) f (u ) g (t u )du F f ( x) * g ( x) F f * g fˆ gˆ j 2ux f ( x ) * g ( x ) e dx f ( y ) g ( x y )e j 2ux dtdy f ( y )e j 2uy G (u )dy F 1 fˆ gˆ f * g f ( y )e j 2uy dyG (u ) F (u )G (u ) fˆ gˆ 25 Convolution Edge detection 26 The Adjoint of the Fourier Transform Theorem: Suppose f and g er are square integrable. Then: F f g Proof: F f g 2 L f F 1g L2 L2 fˆ (u ) g (u )du f ( x )e j 2ux dt g (u )du f ( x) g (u )e j 2ux du dx f ( x) F g ( x)dx 1 f F 1 g L2 27 Plancherel Formel - The Parselval’s Theorem Theorem: Suppose f and g are square integrable. Then: F f F g L2 F 1 f F 1 g Proof: L2 F f F g L2 F 1 f F 1 g 2 L f g f g f F 1 F g F F 1 f g In paricular L2 F[f] L2 f L2 L2 L2 f g f g L2 L2 28 L2 The Rayleigh’s Theorem Conservation of Energy F f F g L2 F[f] L2 f energy f g 2 f ( x) dx L2 f ( x) dx 2 2 L 2 F (u ) du f ( x) dx 2 f ( x) f * ( x)dx The energy of a signal in the time domain is the same as the energy in the frequency domain f fˆ L2 L2 29 The Fourier Series Expansion u a discrete variable - Forward transform Suppose f(t) is a transient function that is zero outside the interval [-T/2,T/2] or is considered to be one cycle of a periodic function. We can obtain a sequence of coefficients by making a discrete variable and integrating only over the interval. fˆ (u ) f ( x)e j 2ux dx fˆn fˆn (n u ) T /2 f ( x)e j 2ux dx T / 2 T /2 f ( x)e jn2u dx T / 2 u 1 T 30 The Fourier Series Expansion u a discrete variable - Inverse transform fˆn fˆn (n u ) T /2 f ( x )e jn 2u 1 u T dx T / 2 The inverse transform becomes: f ( x) fˆ ( x)e j 2ux du n fˆn (n u )e jn2ux u n fˆn e jn 2 x T 2 1 1 ˆ jn T f ne T T n x 31 The Fourier Series Expansion cn coefficients fˆn fˆn (n u ) T /2 f ( x )e T / 2 jn 2u dx 1 u T 2 2 jn x jn x 1 f ( x) fˆ ( x)e j 2ux du fˆn e T cn e T T n n f ( x) c e n jn 2 x T n 2 n j x 1 T cn f ( x )e dx T T / 2 T /2 32 The Fourier Series Expansion zn, an, bn coefficients f ( x) c e n jn 2 x T n 2 n j x 1 T cn f ( x ) e T T/ 2 T /2 f ( x) c e n jn n 2 x T 2 n 2 n j x j x 1 T T f ( x ) e dx e n T T / 2 T /2 2 n 2 n j x j x a0 1 T T f ( x ) e dx e 2 n T T/ 2 T /2 n0 T /2 T /2 2 n 2 n 2 n 2 n j x j x j x j x a0 1 a0 T T T T f ( x )e dx e f ( x )e dx e zn 2 n 1 T T / 2 T / 2 2 n 1 T /2 T /2 2 n 2 n 2 n 2 n 2 nx 2 nx j x j x j t j x j j 1 1 z n f ( x)e T dx e T f ( x)e T dx e T (an ibn )e T (an ibn )e T T T / 2 T / 2 2 2 nx j 2 an ibn f (t )e T T T / 2 T /2 33 The Fourier Series Expansion an,bn coefficients a0 f (t ) z n 2 n 1 j 1 z n (an ibn )e 2 T /2 an ibn 2 nx T 2 f (t )e T T/ 2 j (an ibn )e 2 nx T j 2 nx T a0 2nx 2nx f ( x) an cos bn sin 2 n 1 T T 2 2nx an f ( x ) cos dx T T / 2 T T /2 2 2nx bn f ( x ) sin dx T T / 2 T T /2 34 f ( x) 4 N 1 sin (2i 1) 2i 1 i 1 Fourier Series 2 x Pulse train Pulse train approximated by Fourier Serie N=1 N=2 N=5 N = 10 35 Fourier Series Pulse train – Java program 36 Pulse Train approximated by Fourier Serie f(x) square wave (T=2) a0 2nx 2nx an cos bn sin 2 n 1 T T 4 1 sin[( 2n 1)x] n 1 2n 1 f ( x) f ( x) 4 N 1 2n 1 sin[( 2n 1)x] n 1 N=1 N=2 N=10 37 2 N 1 f ( x) (1)i 1 sin( ikx) k i 1 i Fourier Series k Zig tag Zig tag approximated by Fourier Serie N=1 N=2 N=5 N = 10 38 2 N 1 (-1)i f ( x ) 4 cos(ikx) 2 3 2 i 1 (ik ) 2 Fourier Series k Negative sinus function N=1 Negative sinus function approximated by Fourier Serie N=2 N=5 N = 10 39 2 1 2 N 1 f ( x) sin( kx) cos(2ikx) 2 2 i 1 (2i ) 1 1 Fourier Series k Truncated sinus function Truncated sinus function approximated by Fourier Serie N=1 N=2 N=5 N = 10 40 2 Fourier Series Line Line approximated by Fourier Serie N=1 N a0 N f ( x) a j cos( jkx) b j sin( jkx) 2 j 0 j 0 k L N=2 1 a j f ( x) cos( jkx)dx L L L N=5 41 N = 50 L L 1 b j f ( x) sin( jkx)dx L L N = 10 Fourier Series Java program for approximating Fourier coefficients Approximate functions by adjusting Fourier coefficients (Java program) 42 The Discrete Fourier Transform - DFT Discrete Fourier Transform - Discretize both time and frequency Continuous Fourier transform Discrete frequency Fourier Serie u n u fˆ (u ) T /2 f ( x )e j 2ux T / 2 f ( x) fˆ (u)e j 2ux dx u fˆn fˆ (nu ) Discrete frequency and time Discrete Fourier Transform 1 T t i t t T /2 f ( x )e T / 2 2 du 1 ˆ jn T f ( x) f n e T n x T N n jn 2u dx N /2 j 2 i T N ˆf fˆ (nu ) fi e n N i N / 2 1 ˆ j 2 N n f i f (ix) f n e T n i 43 The Discrete Fourier Transform - DFT Discrete Fourier Transform - Discretize both time and frequency { fi } sequence of length N, taking samples of a continuous function at equal intervals n N /2 j 2 i T N ˆf fˆ (nu ) fi e n N i N / 2 ˆf 1 n N 1 ˆ j 2 N n f i f (ix) f n e T n 1 fi N i N 1 fe i 0 j 2 i N 1 fˆ e n 0 n i N j 2 i n N n 44 Continuous Fourier Transform in two Dimensions Def The Fourier transform of a two-dimentional function f(x,y) fˆ (u, v) f ( x, y )e j 2 (uxvy ) dxdy The Inverse Fourier Transform f ( x, y ) fˆ (u, v)e j 2 (ux vy ) dudv 45 The Two-Dimensional DFT and Its Inverse 1 fˆ (u, v) MN M 1 N 1 f ( x, y)e j 2 ( 1 f ( x, y ) MN M 1 N 1 j 2 ( u v x y ) M N x 0 y 0 fˆ (u, v)e u v x y ) M N x 0 y 0 46 Fourier Transform in Two Dimensions Example 1 47 Fourier Transform in Two Dimensions Example 2 48 End 49