logo

English

이곳의 프로그래밍관련 정보와 소스는 마음대로 활용하셔도 좋습니다. 다만 쓰시기 전에 통보 정도는 해주시는 것이 예의 일것 같습니다. 질문이나 오류 수정은 siseong@gmail.com 으로 주세요. 감사합니다.

FFT (Fast Fourier Transform) 고속 푸리에 변환 예제 소스

by digipine posted Oct 29, 2017
?

Shortcut

PrevPrev Article

NextNext Article

Larger Font Smaller Font Up Down Go comment Print
?

Shortcut

PrevPrev Article

NextNext Article

Larger Font Smaller Font Up Down Go comment Print

#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <math.h>

#define MAX(a,b)    (((a) > (b)) ? (a) : (b))
#define PI  3.141592

typedef struct {
    double real, imag;
} COMPLEX;

   FILE *fp;
 
/* function prototypes for DSP disk file functions */
   FILE *open_read(char *);
   FILE *open_write(char *);
   void read_record(FILE *,double *,int);
   void write_record(double *,int);
   double *read_double_record(FILE *);

/* function prototypes for dft and inverse dft functions */
   void fft(COMPLEX *,int);
   int log2(unsigned int);

/************************************************************************
IFFTTEST.C - Demonstrate and test FFT and Inverse FFT functions
Requires time domain data file in DSP_FILE format.
Generates DSP_FILE format files for spectral magnitude
and reconstructed time domain data.
************************************************************************/

void main(void)
{
    int          i, length, fft_length, m;
    double       tempflt;
    double       *signal,*log_mag;
    COMPLEX      *samp;

// Read the input data file from the dsp format.  데이타파일 읽는 명령
    fp = open_read("test.txt");
    length = 128;
    signal = read_double_record(fp);

// determine fft size and allocate the complex array - fft size와 배열
    m = log2(length);
    fft_length = 1 << m;   //bit shift
    samp = (COMPLEX *) calloc(fft_length, sizeof(COMPLEX));

// copy input signal to complex array and do the fft
    for (i=0; i<length; i++)
  samp[i].real = signal[i];
    fft(samp, m);

/* find log magnitude and store for output */
    log_mag = (double *)calloc(fft_length, sizeof(double));
    for (i=0; i< fft_length; i++)
 {
        tempflt  = samp[i].real * samp[i].real;
        tempflt += samp[i].imag * samp[i].imag;
        tempflt =  sqrt(tempflt);
        log_mag[i] = 10 * log10(MAX(tempflt, 1.e-14));
    }
    fp = open_write("test_out.txt");  //결과 저장 텍스트 //
    write_record(log_mag, length);
}

/**************************************************************************
DISKIO.C - Source code for DSP data format read and write functions
    open_read          open DSP data file to be read
    open_write         create header and open DSP data file for write
    read_record        read one record
    write_record       write one record
    read_float_record  read one record and convert to float array
*************************************************************************/

/***************************************************************************************
open_read - open a DSP data file for read
Returns a pointer to a DSP_FILE structure allocated by the function and opens file_name.
Allocation errors or improper type causes a call to exit(1).
A bad file_name returns a NULL pointer.
DSP_FILE *open_read(char *file_name)
***************************************************************************************/

FILE *open_read(char *file_name) /* file name string */
{
/* allocate the DSP data file structure */

    fp = (FILE *) malloc(sizeof(FILE));

/* open file for text read and update */
    fp = fopen(file_name,"r");
    return(fp);
}

/**************************************************************************
open_write - open a DSP data file for write
Returns a pointer to a DSP_FILE structure allocated by the function.
Allocation errors or improper type causes a call to exit(1).
A bad file name returns a NULL pointer.
DSP_FILE *open_write(char *file_name, int records,int rec_len)
    file_name       pointer to file name string
    records         number of records of data to be written
    rec_len         number of elements in each record
*************************************************************************/

FILE *open_write(char * file_name)
{
/* allocate the DSP data file structure */
    fp = (FILE *) malloc(sizeof(FILE));

/* open file for text write and update*/
    fp = fopen(file_name,"wt");
    return(fp);
}

/********************************************************************
read_record - read one record of DSP data file
Exits if a read error occurs or if the DSP_FILE structure is invalid.   
Void read_record(FILE *fp , double *input , int length)
********************************************************************/

void read_record(FILE *fp ,double * input , int length)
{
    int status , i;
 for(i=0 ; i < length ; i++)
 {
  status = fscanf(fp,"%lf\n", &input[length + i]);
    }
}

/************************************************************************************************************************
read_double_record - read one record of DSP data file and convert to float array of values.
Returns a pointer to the beginning of the allocated float array of values representing the record read from the DSP_FILE.
Exits if a read or allocation error occurs.
float *read_double_record(DSP_FILE *dsp_info)
************************************************************************************************************************/

double *read_double_record(FILE * fp)
{
    static double *buf;            /* input buffer to read data in */
    double *out;                    /* return output pointer */
    double *out_ptr;
 double *d_ptr;
    int i,length,length_in;

  length = 128;
    length_in = 256;
 
    buf = (double *) calloc(length_in,sizeof(double));
    out = (double *) calloc(length,sizeof(double));

/* read the record into buf */
    read_record(fp , buf , length);

/* perform conversion to floating point */
     out_ptr = out; 
     d_ptr = buf + 128;
     for(i = 0 ; i < length ; i++)
         *out_ptr++ = (double)(*d_ptr++);
 
  return(out);                 /* return converted pointer */
}

/**************************************************************************
write_record - write one record of DSP_FILE data
Exits if write error occurs or if the DSP_FILE structure is invalid.
void write_record(char *ptr,DSP_FILE *dsp_info, int length)
    ptr        pointer to data to write to disk (type in dsp_info)
    dsp_info   pointer to DSP data file structure   
*************************************************************************/

void write_record(double *ptr,int length)
{
    int status, i;
 for(i=0 ; i < length ; i++)
 {
  status = fprintf(fp,"%lf\n", *ptr++);
    }
}

/**********************************************************************
DFT.C - SOURCE CODE FOR DISCRETE FOURIER TRANSFORM FUNCTIONS
fft     In-place radix 2 decimation in time FFT
log2    Base 2 logarithm
***********************************************************************/

/**********************************************************************
fft - In-place radix 2 decimation in frequency FFT
Requires pointer to complex array, x and power of 2 size of FFT, m
(size of FFT = 2^m).  Places FFT output on top of input COMPLEX array.
void fft(COMPLEX *x, int m)
***********************************************************************/

void fft(COMPLEX *x, int m)

  static COMPLEX *w;           // used to store the w complex array
    static int mstore = 0;       // stores m for future reference
    static int n = 1;            // length of fft stored for future

    COMPLEX u, temp, tm;
    COMPLEX *xi, *xip, *xj, *wptr;

    int i, j, k, l, le, windex;

    double arg, w_real, w_imag, wrecur_real, wrecur_imag, wtemp_real;

    if(m != mstore)
 {
  // free previously allocated storage and set new m
        if(mstore != 0) free(w);
        mstore = m;
        if(m == 0) return;       // if m=0 then done

  // n = 2^m = fft length
        n = 1 << m;  
        le = n/2;  //difference between the upper and lower leg indices

  // allocate the storage for w
        w = (COMPLEX *) calloc(le-1,sizeof(COMPLEX));

  // calculate the w values recursively 계산
        arg = PI/le;         //  PI/le calculation
        wrecur_real = w_real = cos(arg);
        wrecur_imag = w_imag = -sin(arg);
        xj = w;
        for (j = 1 ; j < le ; j++)
  {
            xj->real = (double)wrecur_real;
            xj->imag = (double)wrecur_imag;
            xj++;
            wtemp_real = wrecur_real*w_real - wrecur_imag*w_imag;
            wrecur_imag = wrecur_real*w_imag + wrecur_imag*w_real;
            wrecur_real = wtemp_real;
        }
    }
 // start fft 푸리에 변환 시작
    le = n;
    windex = 1;
    for (l = 0 ; l < m ; l++)
 {
        le = le/2;
  // first iteration with no multiplies
        for(i = 0 ; i < n ; i = i + 2*le)
  {
            xi = x + i;
            xip = xi + le;
            temp.real = xi->real + xip->real;
            temp.imag = xi->imag + xip->imag;
            xip->real = xi->real - xip->real;
            xip->imag = xi->imag - xip->imag;
            *xi = temp;
        }
  // remaining iterations use stored w
        wptr = w + windex - 1;
        for (j = 1 ; j < le ; j++)
  {
            u = *wptr;
            for (i = j ; i < n ; i = i + 2*le)
   {
                xi = x + i;
                xip = xi + le;
                temp.real = xi->real + xip->real;
                temp.imag = xi->imag + xip->imag;
                tm.real = xi->real - xip->real;
                tm.imag = xi->imag - xip->imag;            
                xip->real = tm.real*u.real - tm.imag*u.imag;
                xip->imag = tm.real*u.imag + tm.imag*u.real;
                *xi = temp;
            }
            wptr = wptr + windex;
        }
        windex = 2*windex;
    }
 // rearrange data by bit reversing
    j = 0;
    for (i = 1 ; i < (n-1) ; i++)
 {
        k = n/2;
        while(k <= j)
  {
            j = j - k;
            k = k/2;
        }
        j = j + k;
        if (i < j)
  {
            xi = x + i;
            xj = x + j;
            temp = *xj;
            *xj = *xi;
            *xi = temp;
        }
    }
}

/**************************************************************************
log2 - base 2 logarithm
Returns base 2 log such that i = 2^ans where ans = log2(i).
if log2(i) is between two values, the larger is returned.
int log2(unsigned int x)
*************************************************************************/
int log2(unsigned int x)
{
    unsigned int mask,i;

    if(x == 0) return(-1);      // zero is an error, return -1
    x--;                        // get the max index, x-1
    for(mask = 1 , i = 0 ; ; mask *= 2 , i++)
 {
        if(x == 0) return(i);   // return log2 if all zero
        x = x & (~mask);        // AND off a bit
    }

}


http://starpotato.springnote.com/pages/3236324 권영준님의 노트에서...

Jan.06, 2009 : FFT의 개괄적인 이해는 2d fast fourier transform 을 참조하세요

FFT 소스를 요청하는 분들이 많이 계셔서 소스코드를 공개합니다. 이게 기본틀은 어디서 코드를 분석한 것이라서 거의 똑같다고 보시면 됩니다. 간단한 주석을 달았구요. 도움이 되시길 빕니다.

          /* File            :       struct.h
           * Creator      :       BlackEngine
           * Date         :       2006/03/28
           * Version      :       0.0.1 ( created )
           * Descript     :       Definition of Complex Number , and etc
           */
           
          typedef struct _COMPLEX {
                  double     real;
                  double     imag;
          } COMPLEX, *pCOMPLEX;
          
          /* File            :       fft.h
           * Creator      :       BlackEngine
           * Date         :       2006/03/28
           * Version      :       0.0.1 ( created )
           * Descript     :       Perform the FFT Algorithm
           */
          #ifndef _BLACKENGINE_FFT_H_
          #define _BLACKENGINE_FFT_H_
           
          #include "struct.h"
          /* TODO : at 03/29 , decision the return values of FFT Function. */
          int fft_2d(COMPLEX ** ,int ,int ,int );
          int fft(int ,int ,double *,double *);
          #endif /* ifndef _BLACKENGINE_FFT_H_ */
          
          /* File            :       fft.c
           * Creator      :       BlackEngine
           * Date         :       2006/03/28
           * Version      :       0.0.1 ( created )
           * Descript     :       Perform the FFT Algorithm ( 2-D Image FFT )
           */
          #include 
          #include 
          #include
           
          #include "fft.h"
           
          #define TRUE 1
          #define FALSE 0
          /*#define NULL 0*/
           
          /* Powerof2 function : compare the integer with 2^m
           * and find maximum m , that is 2^m &lt;= nx
           * twopm = 2^m
           */
          int Powerof2(int nx, int* m, int* twopm) {
                  int pwr;
                  *m = 0;
                  for ( pwr = 1 ; pwr &lt; nx ; pwr = pwr*2) {
                          *m = *m + 1;
                  }
                  *twopm = pwr;
                  return(TRUE);
          }
           
          int fft_2d(COMPLEX **c, int nx, int ny, int dir) {
                  int i,j;
                  int m,twopm;
                  double *real,*imag;
           
                  /* Transform the rows */
                  real = (double *)malloc(nx * sizeof(double));
                  imag = (double *)malloc(nx * sizeof(double));
                  if (real == NULL || imag == NULL)
                          return(FALSE);
                  if (!Powerof2(nx,&amp;m,&amp;twopm) || twopm != nx)
                          return(FALSE);
                  for (j=0;j&lt;&lt;= m;
           
                  /* N의 절반값을 half에 저장한다. */
                  half = N &gt;&gt; 1;
           
                  /* Bit Reverse를 하는 과정 */
                  for ( i = 0 , j = 0 ; i &lt; N - 1 ; i++ ) {
                          if ( i &lt; j )
                          {
                                  swap_x = x[i];
                                  swap_y = y[i];
                                  x[i] = x[j];
                                  y[i] = y[j];
                                  x[j] = swap_x;
                                  y[j] = swap_y;
                          }
                          step = half;
           
                          while ( step &lt;= j )
                          {
                                  j = j - k;
                                  k &gt;&gt;= 1;
                          }
                          j = j + step;
                  }
           
                  /* 초기의 값을 입력하는 부분,
                   * factor의 역할은 각 단계(Stage)에서 omega의 값을 변경하기 위한
                   * 기본 값이다. 즉 omega N= 4일 경우에는 각도가 pi/2 만큼씩 변하므로
                   * exp(-j pi/2) 의 값을 가지고 있는 것이 factor이다
                   * factor는 각 stage가 올라갈 때 마다 변경되어야 한다.
                   * 그 각도가 절반으로 줄어야 하므로
                   * cos(t) + j sin(t) 가 cos(t/2) + jsin(t/2) 로 되어야 한다.
                   * cos(t) = 2cos^2(t/2) -1 이므로 cos(t/2) = sqrt((1+cos(t))/2)이고
                   * sin(t/2) = sqrt(1-cos^2(t/2))이므로 sqrt((1-cos(t))/2) 이다.
                   */
                  factor_x = -1.0;
                  factor_y = 0;
                  butterfly = 1;
           
                  for ( stage = 0 ; stage &lt; m ; stage++ )
                  {
                          /* step 은 butterfly가 두개의 샘플을 가지고 작업하므로
                           * 현재 위치에서 몇번째 샘플을 가지고 해야 하는 가를
                           * 가리키는 변수이다.
                           * butterfly는 같은 omega 값에의해 반복되는 butterfly가
                           * 존재하므로 그 위치를 가리키기 위한 값이다.
                           */
                          step = butterfly;
                          butterfly &lt;&lt;= 1;
                          /* 각 단계에서 omega는 항상 1+j0으로 시작한다. */
                          omega_x = 1.0;
                          omega_y = 0;
                          for ( numofbutter = 0 ; numofbutter &lt; step ; numofbutter++)
                          {
                                  for( i = numofbutter; i &lt; N; i = i + butterfly)
                                  {
                                          /* omega_N(k) 일때 omega_N(k+N/2) = -omega_N(k)
                                           * 의 성질을 이용한다. 그래서 두번째 샘플에 미리
                                           * omega_N(k)를 곱한 후 결과의 처음 샘플에는
                                           * 그냥 더하고 두번째 샘플에는 빼준다(-이므로)
                                           */
                                          swap_x = x[i+step]* omega_x - y[i+step]*omega_y;
                                          swap_y = y[i+step]*omega_x + x[i+step]*omega_y;
           
                                          x[i+step] = x[i] - swap_x;
                                          y[i+step] = y[i] - swap_y;
           
                                          x[i] = x[i] + swap_x;
                                          y[i] = y[i] + swap_y;
           
                                  }
                                  /* 공통된 omega를 사용하는 샘플을 다 곱한 후
                                   * omega를 factor의 각도만큼 증가시켜야 한다.
                                   * omega(k+1) 의 값은 omega와 factor의 곱을 통해
                                   * 구할 수 있다.
                                   */
                                  swap_x = omega_x * factor_x - omega_y * factor_y;
                                  swap_y = omega_y * factor_x + omega_x * factor_y;
                                  omega_x = swap_x;
                                  omega_y = swap_y;
           
                          }
                          /* 한 단계가 끝났으므로 factor를 기존 값에서 각도가
                           * 절반인 값으로 구해야 한다.
                           */
                          factor_y = sqrt((1.0+factor_x)/2.0);
                          if( dir == 1)
                                  factor_y = -factor_y;
                          factor_x = sqrt((1.0-factor_x)/2.0);
                  }
                  /* Inverse FFT 의 경우에는 sampling을 해줘야 한다.
                   * 다른 소스에서는 FFT의 값이 크게 나오므로 fft시에 sampling을
                   * 하고 inverse에서 안하는 경우도 있다 ( 상관없음 )
                   */
                  if( dir == -1 )
                  {
                          for ( i = 0 ; i &lt; N ; i++)
                          {
                                  x[i] /= N;
                                  y[i] /= N;
                          }
                  }
           
                  return TRUE;
          }
          


=======dft.h============

typedef struct {
	double real, imag;
} COMPLEX;
#define PI 3.141592

void fft(COMPLEX *,int);
void ifft(COMPLEX *,int);
void dft(COMPLEX *,COMPLEX *,int);
void idft(COMPLEX *,COMPLEX *,int);
void rfft(double *,COMPLEX *,int);
void ham(COMPLEX *,int);
void han(COMPLEX *,int);
void triang(COMPLEX *,int);
void black(COMPLEX *,int);
void harris(COMPLEX *,int);
int log2(unsigned int);


=========dft.c==========

#include <math.h>
#include <stdlib.h>
#include <stdio.h>
#include "dft.h"

void fft(COMPLEX *x, int m)
{ 
	static COMPLEX *w; 
	static int mstore = 0;
	static int n = 1;
	COMPLEX u,temp,tm;
	COMPLEX *xi,*xip,*xj,*wptr;
	int i,j,k,l,le,windex;
	double arg,w_real,w_imag,wrecur_real,wrecur_imag,wtemp_real;
	if(m != mstore) {

		if(mstore != 0) free(w);
		mstore = m;
		if(m == 0) return; 
		n = 1 << m; 
		le = n/2; 

		w = (COMPLEX *) calloc(le-1,sizeof(COMPLEX));
		if(!w) {
			printf("\nUnable to allocate complex W array\n");
			exit(1);
		}

		arg = PI/le; 
		wrecur_real = w_real = cos(arg);
		wrecur_imag = w_imag = -sin(arg);
		xj = w;
		for (j = 1 ; j < le ; j++) {
			xj->real = (double)wrecur_real;
			xj->imag = (double)wrecur_imag;
			xj++;
			wtemp_real = wrecur_real*w_real - wrecur_imag*w_imag;
			wrecur_imag = wrecur_real*w_imag + wrecur_imag*w_real;
			wrecur_real = wtemp_real;
		}
	}

	le = n;
	windex = 1;
	for (l = 0 ; l < m ; l++) {
		le = le/2;

		for(i = 0 ; i < n ; i = i + 2*le) {
			xi = x + i;
			xip = xi + le;
			temp.real = xi->real + xip->real;
			temp.imag = xi->imag + xip->imag;
			xip->real = xi->real - xip->real;
			xip->imag = xi->imag - xip->imag;
			*xi = temp;
		}


		wptr = w + windex - 1;
		for (j = 1 ; j < le ; j++) {
			u = *wptr;
			for (i = j ; i < n ; i = i + 2*le) {
				xi = x + i;
				xip = xi + le;
				temp.real = xi->real + xip->real;
				temp.imag = xi->imag + xip->imag;
				tm.real = xi->real - xip->real;
				tm.imag = xi->imag - xip->imag; 
				xip->real = tm.real*u.real - tm.imag*u.imag;
				xip->imag = tm.real*u.imag + tm.imag*u.real;
				*xi = temp;
			}
			wptr = wptr + windex;
		}
		windex = 2*windex;
	} 

	j = 0;
	for (i = 1 ; i < (n-1) ; i++) {
		k = n/2;
		while(k <= j) {
			j = j - k;
			k = k/2;
		}
		j = j + k;
		if (i < j) {
			xi = x + i;
			xj = x + j;
			temp = *xj;
			*xj = *xi;
			*xi = temp;
		}
	}
}

main(){
	int i;
	COMPLEX com[1];
	fft(com,1);
	for(i=0;i<sizeof(com);i++)
		printf("%f\n",com[i].real);
}
 
TAG •

List of Articles
No. Subject Author Date Views
18 [Linux, OSX] pfctl - Packet FIlter Control 사용법 digipine 2017.11.02 2552
17 Xcode 없이 맥에 '명령어 라인 도구(Command Line Tools)'를 설치하는 방법 엉뚱도마뱀 2018.12.26 2577
16 [iOS] Bluetooth로 App을 백그라운드 모드로 실행는 방법 lizard2019 2020.02.11 2786
15 OpenAL 사용법 정리 1 digipine 2017.11.01 3189
14 난수발생기 개론 엉뚱도마뱀 2017.11.22 3709
13 비밀번호 해쉬에 Salt(소금) 첨가하기 file 엉뚱도마뱀 2017.11.23 3982
12 공짜 무료 C/C++ 컴파일러들 file digipine 2017.10.28 4154
11 ATmega8 MCU 간의 TWI 기능을 이용한 I2C 통신 digipine 2017.11.02 4877
10 WinPCap과 Ethereal, Wireshark 을 이용한 스니핑(Sniffing) digipine 2017.10.29 5301
9 Phabricator 설치 가이드 우분투 16.04 기준 digipine 2017.11.02 6058
8 Photoshop CC 2018 한글 영문 변환 언어팩, 포토샵 언어변경 file 엉뚱도마뱀 2018.07.04 7531
7 Linux init.d 에서 등록하기. 부팅 시 자동실행 설정 digipine 2017.11.03 9124
» FFT (Fast Fourier Transform) 고속 푸리에 변환 예제 소스 digipine 2017.10.29 11861
5 [Swift, MacOS] 맥 한글 파일명이 윈도우에서 자소 분리되는 현상 해결, NFD, NFC 엉뚱도마뱀 2018.12.11 14098
4 [Linux] ubuntu 16.04에 QT Creator 설치하기 digipine 2017.11.02 23934
3 What is Android Repo? and Manual, Download file digipine 2017.11.02 84016
2 언어 IDE 별로 git ignore 파일을 자동으로 만들어 주는 사이트 엉뚱도마뱀 2018.12.17 102293
1 TCP/IP State Transition - TCP 스택 포팅 시 참조 file digipine 2017.11.02 152983
Board Pagination Prev 1 2 3 4 5 Next
/ 5