## Fourier Series

### About Fourier Series Models

The Fourier series is a sum of sine and cosine functions that describes a periodic signal. It is represented in either the trigonometric form or the exponential form. The toolbox provides this trigonometric Fourier series form

$$y={a}_{0}+{\displaystyle \sum _{i=1}^{n}{a}_{i}\mathrm{cos}(iwx)+{b}_{i}\mathrm{sin}(iwx)}$$

where *a*_{0} models a
constant (intercept) term in the data and is associated with the *i* =
0 cosine term, *w* is the fundamental frequency of
the signal, *n* is the number of terms (harmonics)
in the series, and 1 ≤ *n* ≤ 8.

For more information about the Fourier series, refer to Fourier Analysis and Filtering.

### Fit Fourier Models Interactively

Open the Curve Fitting app by entering

`cftool`

. Alternatively, click Curve Fitting on the Apps tab.In the Curve Fitting app, select curve data (

**X data**and**Y data**, or just**Y data**against index).Curve Fitting app creates the default curve fit,

`Polynomial`

.Change the model type from

`Polynomial`

to`Fourier`

.

You can specify the following options:

Choose the number of terms:

`1`

to`8`

.Look in the

**Results**pane to see the model terms, the values of the coefficients, and the goodness-of-fit statistics.(Optional) Click

**Fit Options**to specify coefficient starting values and constraint bounds, or change algorithm settings.The toolbox calculates optimized start points for Fourier series models, based on the current data set. You can override the start points and specify your own values in the Fit Options dialog box.

For more information on the settings, see Specifying Fit Options and Optimized Starting Points.

For an example comparing the library Fourier fit with custom equations, see Custom Nonlinear ENSO Data Analysis.

### Fit Fourier Models Using the fit Function

This example shows how to use the `fit`

function to fit a Fourier model to data.

The Fourier library model is an input argument to the `fit`

and `fittype`

functions. Specify the model type `fourier`

followed by the number of terms, e.g., `'fourier1'`

to `'fourier8'`

.

This example fits the El Nino-Southern Oscillation (ENSO) data. The ENSO data consists of monthly averaged atmospheric pressure differences between Easter Island and Darwin, Australia. This difference drives the trade winds in the southern hemisphere.

The ENSO data is clearly periodic, which suggests it can be described by a Fourier series. Use Fourier series models to look for periodicity.

**Fit a Two-Term Fourier Model**

Load some data and fit an two-term Fourier model.

load enso; f = fit(month,pressure,'fourier2')

f = General model Fourier2: f(x) = a0 + a1*cos(x*w) + b1*sin(x*w) + a2*cos(2*x*w) + b2*sin(2*x*w) Coefficients (with 95% confidence bounds): a0 = 10.63 (10.23, 11.03) a1 = 2.923 (2.27, 3.576) b1 = 1.059 (0.01593, 2.101) a2 = -0.5052 (-1.086, 0.07532) b2 = 0.2187 (-0.4202, 0.8576) w = 0.5258 (0.5222, 0.5294)

plot(f,month,pressure)

The confidence bounds on `a2`

and `b2`

cross zero. For linear terms, you cannot be sure that these coefficients differ from zero, so they are not helping with the fit. This means that this two term model is probably no better than a one term model.

**Measure Period**

The `w`

term is a measure of period. `2*pi/w`

converts to the period in months, because the period of `sin()`

and `cos()`

is `2*pi`

.

w = f.w

w = 0.5258

2*pi/w

ans = 11.9497

`w`

is very close to 12 months, indicating a yearly period. Observe this looks correct on the plot, with peaks approximately 12 months apart.

**Fit an Eight-Term Fourier Model**

`f2 = fit(month,pressure,'fourier8')`

f2 = General model Fourier8: f2(x) = a0 + a1*cos(x*w) + b1*sin(x*w) + a2*cos(2*x*w) + b2*sin(2*x*w) + a3*cos(3*x*w) + b3*sin(3*x*w) + a4*cos(4*x*w) + b4*sin(4*x*w) + a5*cos(5*x*w) + b5*sin(5*x*w) + a6*cos(6*x*w) + b6*sin(6*x*w) + a7*cos(7*x*w) + b7*sin(7*x*w) + a8*cos(8*x*w) + b8*sin(8*x*w) Coefficients (with 95% confidence bounds): a0 = 10.63 (10.28, 10.97) a1 = 0.5668 (0.07981, 1.054) b1 = 0.1969 (-0.2929, 0.6867) a2 = -1.203 (-1.69, -0.7161) b2 = -0.8087 (-1.311, -0.3065) a3 = 0.9321 (0.4277, 1.436) b3 = 0.7602 (0.2587, 1.262) a4 = -0.6653 (-1.152, -0.1788) b4 = -0.2038 (-0.703, 0.2954) a5 = -0.02919 (-0.5158, 0.4575) b5 = -0.3701 (-0.8594, 0.1192) a6 = -0.04856 (-0.5482, 0.4511) b6 = -0.1368 (-0.6317, 0.3581) a7 = 2.811 (2.174, 3.449) b7 = 1.334 (0.3686, 2.3) a8 = 0.07979 (-0.4329, 0.5925) b8 = -0.1076 (-0.6037, 0.3885) w = 0.07527 (0.07476, 0.07578)

plot(f2,month,pressure)

**Measure Period**

w = f2.w

w = 0.0753

(2*pi)/w

ans = 83.4736

With the `f2`

model, the period `w`

is approximately 7 years.

**Examine Terms**

Look for the coefficients with the largest magnitude to find the most important terms.

`a7`

and`b7`

are the largest. Look at the`a7`

term in the model equation:`a7*cos(7*x*w)`

.`7*w`

== 7/7 = 1 year cycle.`a7`

and`b7`

indicate the annual cycle is the strongest.Similarly,

`a1`

and`b1`

terms give 7/1, indicating a seven year cycle.`a2`

and`b2`

terms are a 3.5 year cycle (7/2). This is stronger than the 7 year cycle because the`a2`

and`b2`

coefficients have larger magnitude than a1 and b1.`a3`

and`b3`

are quite strong terms indicating a 7/3 or 2.3 year cycle.Smaller terms are less important for the fit, such as

`a6`

,`b6`

,`a5`

, and`b5`

.

Typically, the El Nino warming happens at irregular intervals of two to seven years, and lasts nine months to two years. The average period length is five years. The model results reflect some of these periods.

**Set Start Points**

The toolbox calculates optimized start points for Fourier fits, based on the current data set. Fourier series models are particularly sensitive to starting points, and the optimized values might be accurate for only a few terms in the associated equations. You can override the start points and specify your own values.

After examining the terms and plots, it looks like a 4 year cycle might be present. Try to confirm this by setting `w`

. Get a value for `w`

, where 8 years = 96 months.

w = (2*pi)/96

w = 0.0654

Find the order of the entries for coefficients in the model ('f2') by using the `coeffnames`

function.

coeffnames(f2)

`ans = `*18x1 cell*
{'a0'}
{'a1'}
{'b1'}
{'a2'}
{'b2'}
{'a3'}
{'b3'}
{'a4'}
{'b4'}
{'a5'}
{'b5'}
{'a6'}
{'b6'}
{'a7'}
{'b7'}
{'a8'}
{'b8'}
{'w' }

Get the current coefficient values.

coeffs = coeffvalues(f2)

`coeffs = `*1×18*
10.6261 0.5668 0.1969 -1.2031 -0.8087 0.9321 0.7602 -0.6653 -0.2038 -0.0292 -0.3701 -0.0486 -0.1368 2.8112 1.3344 0.0798 -0.1076 0.0753

Set the last coefficient, `w`

, to 0.065.

coeffs(:,18) = w

`coeffs = `*1×18*
10.6261 0.5668 0.1969 -1.2031 -0.8087 0.9321 0.7602 -0.6653 -0.2038 -0.0292 -0.3701 -0.0486 -0.1368 2.8112 1.3344 0.0798 -0.1076 0.0654

Set the start points for coefficients using the new value for `w`

.

f3 = fit(month,pressure,'fourier8', 'StartPoint', coeffs);

Plot both fits to see that the new value for `w`

in `f3`

does not produce a better fit than `f2`

.

plot(f3,month,pressure) hold on plot(f2, 'b') hold off legend( 'Data', 'f3', 'f2')

**Find Fourier Fit Options**

Find available fit options using `fitoptions(modelname)`

, where `modelname`

is the model type `fourier`

followed by the number of terms, e.g., `'fourier1'`

to `'fourier8'`

.

`fitoptions('fourier8')`

ans = Normalize: 'off' Exclude: [] Weights: [] Method: 'NonlinearLeastSquares' Robust: 'Off' StartPoint: [1x0 double] Lower: [1x0 double] Upper: [1x0 double] Algorithm: 'Trust-Region' DiffMinChange: 1.0000e-08 DiffMaxChange: 0.1000 Display: 'Notify' MaxFunEvals: 600 MaxIter: 400 TolFun: 1.0000e-06 TolX: 1.0000e-06

If you want to modify fit options such as coefficient starting values and constraint bounds appropriate for your data, or change algorithm settings, see the options for NonlinearLeastSquares on the `fitoptions`

reference page.

## See Also

`fit`

| `fittype`

| `fitoptions`