FFT iOS的值增加了兩倍？

你好我運行FFT功能在iOS平臺的模擬器FFT iOS的值增加了兩倍？

像這樣

-(Matrix*) spectralDensityWithVec:(Matrix*) vector { 

    NSLog(@"Vector:%@",vector); 
    int sampleSize = [vector rows]; 

    double *r = [vector array]; // arc will handle the memory 

    float * f = (float*) malloc(sampleSize * sizeof(float)); 

    for (int i = 0;i< sampleSize;i++) { 
     f[i] = r[i]; 
    } 

    // Working with 128 samples which is 2^7 
    // 7 
    vDSP_Length log2n = log2([self nextPowerOf2WithNumber:sampleSize]); 

    FFTSetup fftSetup = vDSP_create_fftsetup(log2n,FFT_RADIX2); 

    int nOver2 = sampleSize/2; 

    COMPLEX_SPLIT A; 

    A.realp = (float *) malloc(nOver2*sizeof(float)); 
    A.imagp = (float *) malloc(nOver2*sizeof(float)); 

    vDSP_ctoz((COMPLEX*)f, 2, &A, 1, nOver2); 

    vDSP_fft_zrip(fftSetup, &A,1 ,log2n, FFT_FORWARD); 

    Matrix* PSD = [Matrix matrixOfRows:sampleSize/2 Columns:1]; 

    Matrix* fftResult = [Matrix matrixOfRows:sampleSize Columns:1]; 

    // obtain the imaginary/real parts 
    for (int i = 0;i < sampleSize/2;i++) { 
     float realp = A.realp[i]; 
     float imagp = A.imagp[i]; 

     NSLog(@"real:%g img:%g",realp,imagp); 

     float val = sqrtf((realp * realp) + (imagp * imagp)) * 2.0f /(float)sampleSize ; 

     [PSD setValue:val Row:i Column:0]; 
     [fftResult setValue:realp Row:i Column:0]; 
     [fftResult setValue:imagp Row:i*2+1 Column:0]; 
    } 

    NSLog(@"FFT Result KW Energy: %@",fftResult); 

    vDSP_destroy_fftsetup(fftSetup); 

    // release memory 
    if (f) { 
     free(f); 
     f = NULL; 
    } 
    if (A.realp) { 
     free(A.realp); 
     A.realp = NULL; 
    } 
    if (A.imagp) { 
     free(A.imagp); 
     A.imagp = NULL; 
    } 

    return PSD; 
}

用這個載體，我得到的

上面的矢量結果的結果：

如何與numpy fft我得到這個結果：

[65.22805000000001, 
0.0, 
-0.0085759487914117867, 
-0.015603049865923341, 
-0.0069788925854868157, 
0.015041138788651551, 
-0.010263059930525575, 
0., 
-0.021832981634872219, 
-0.026161796455543486, 
-0.0018237156004342254, 
-0.013884928458450145, 
-0.0091755911990841002, 
-0.016388522893239089, 
-0.003204667797210365, 
-0.033305086561749964, 
-0.00048255506211134121, 
-0.015783138731002323, 
-0.0075337096722261337, 
-0.0071538868820418761, 
-0.00066829905037151319, 
-0.017467879005406146, 
0.0058655519004814438, 
0.011289870623693908, 
-0.021329700649694319, 
0.0018507094402767038, 
-0.0062391410740094827, 
-0.013248826390988894, 
0.0091139373764349274, 
-0.0076327732856707516, 
-0.003359005157638993, 
0.0064289012241005097, 
-0.013430000000003162, 
-0.0029959999999995546, 
-0.0065307135571379222, 
-0.006605572447413598, 
-0.0099422652522889073, 
-0.015080331241097611, 
-0.0016757131155144375, 
-0.010765855482435674, 
-0.0028346121922537357, 
-0.014091061276246888, 
-0.0014857985251928623, 
-0.0056375351611077946, 
-0.011988911430201157, 
-0.0049400354875454924, 
-0.0099028010321836699, 
-0.0045315332341146703, 
-0.018321444937884485, 
-0.0050991387310010752, 
-0.027683215530298805, 
0.0015174037330767526, 
-0.0075530851759402738, 
0.019037454812601148, 
0.00020298122212324523, 
0.0081699113196296563, 
-0.0011547055231802498, 
0.0043124328279300489, 
-0.0014283591672329676, 
0.0029302575544578394, 
-0.001462892683058185, 
0.0018357150212062624, 
-0.0014666841715850814, 
0.00088237880086386444, 
-0.0014699999999976399, 
0.0]

我試過多種東西，而且我沒有任何想法任何人都知道我可能做錯了什麼？我從來沒有在ios上使用過FFT函數，在檢查文檔之前，它似乎對這種情況沒有幫助。

來源

2015-11-01 George Host

斑點。

比例因子2 is a property of Apple's real FFT implementation。不知道爲什麼這樣，也許它更有效率。

來源

2015-11-01 21:07:45

是的，我剛剛閱讀文檔，如果我將虛構和真實部分除以2，它應該給我相同的結果。然而，有2 0和蘋果表示，他們做出了暗示。第一對和最後一對的兩個虛部爲0。當我在計算中使用矢量時，這會有所幫助嗎？ –

蘋果包真正傅立葉變換：https://developer.apple.com/library/ios/documentation/Performance/Conceptual/vDSP_Programming_Guide/UsingFourierTransforms/UsingFourierTransforms.html所以是的，你必須考慮到這一點時，執行計算。 –

FFT iOS的值增加了兩倍？

回答

相關問題