# How to get the frequency information of an mp3 audio?

I have:

music = Import["http://www.qlcoder.com/uploads/145425353234642.mp3", "mp3"]


Then I get: But what I want to get is F(t)=x Hz(means in t, the frequency is x Hz). Then how to get that?

• What if there are superimposed sounds of multiple frequencies (as is almost always the case)? Then a simple frequency(t) doesn't make sense. You may want to compute a spectrogram: mathematica.stackexchange.com/questions/4017/… – Szabolcs Feb 1 '16 at 13:10
• If you know for sure that there is only a single (possibly amplitude and frequency modulated) sine oscillator active (e.g. $f(t)=a(t)sin(\omega(t)),\omega(t)=\omega_0 + \int_0^t 2\pi f(s) ds$), then you can see your waveform as the real part of a complex (as in complex numbers) oscillator. You can then use a Hilbert transform (see here) to get the corresponding imaginary part of that oscillator and use this to reconstruct $a(t)$ and $f(t)$ (which is just $\frac{d\omega(t)}{dt}$). – Thies Heidecke Feb 1 '16 at 18:03

Duration = 9.95265 sec, frequency in Hz on x axis:

music = Import["http://www.qlcoder.com/uploads/145425353234642.mp3", "mp3"];
amps = music[[1, 1, 1, All]];
namps = Length@amps; (* namps = 438912 corresponds to 9.95265 sec *)

sr = 44100; (* your sampling rate in Hz *)
inc = sr/namps; (* increment *)
freq = Table[f, {f, 0, sr - inc, inc}] // N;

y = Abs@Fourier[amps, FourierParameters -> {-1, 1}];
data = Transpose[{freq, y}];

fmax = 3000; (* you can go up to Nyquist frequence = max frequency = 22050 Hz *)

ListLinePlot[data, Frame -> True, Joined -> True,
PlotStyle -> {RGBColor[0, 0, 1], Thickness[0.002]},
FrameLabel -> {{"Amplitude", ""}, {"f (Hz)", ""}},
BaseStyle -> {FontWeight -> "Bold", FontSize -> 20,
FontFamily -> "Calibri"}, PlotRange -> {{0, fmax}, All},
ImageSize -> 800] • Is it possible to get the peak frequency value data somehow? – Dheeraj Kumar Jan 24 '19 at 18:22

You can see the spectrum of the first note played, (first 40000 points)

ListLogLogPlot[
{#, # PeakDetect[#, 5, 10^-2]} &@
Abs@Fourier@music[[1, 1, 1, 1 ;; 40000]]
, Joined -> {True, False}
, PlotStyle -> {Gray, Red}
, Filling -> Axis
, PlotRange -> {{100, 1000}, All}
, PlotTheme -> "Scientific"] But beware that the scaling is not in Hertz here To get the scaling correct use:

sft[d_, sr_] := Block[{n, ft, fy},
n = Length[d];
fy = Take[N@Abs[Fourier[d]], n/2];
ft = N@Range[0, n/2 - 1] sr/n;
SortBy[First]@Transpose[{ft, fy}]
]

ListLogLogPlot[
{#, Part[#,
Flatten@Position[PeakDetect[#[[All, 2]], 5, 10^-2], 1]]} &@
sft[music[[1, 1, 1, 1 ;; 40000]], music[[1, 2]]]
, Joined -> {True, False}
, PlotStyle -> {Gray, Red}
, Filling -> Axis
, PlotRange -> {{100, 1000}, All}
, PlotTheme -> "Scientific"] To see the peaks of the first two notes:

Part[#, Flatten@Position[PeakDetect[#[[All, 2]], 5, 10^-2], 1]] &@
sft[music[[1, 1, 1, 1 ;; 40000]], music[[1, 2]]]

 {{174.195, 4.61849}, {350.595, 5.65904}, {524.79, 3.73059}}

Part[#, Flatten@Position[PeakDetect[#[[All, 2]], 5, 10^-2], 1]] &@
sft[music[[1, 1, 1, 40001 ;; 70000]], music[[1, 2]]]

{{164.64, 3.93114}, {330.75, 3.66594}, {495.39, 1.70307}, {660.03, 4.29606}, {826.14, 1.9469}}