Incorrect downsampling audio range

Hi, I have an .wav input audio with the signal in the range [-1;1] at 44.1kHz and I try to apply a down-sampling to about 2kHz but as result I have a signal that exceed the original audio range [-1;1]. It seen like the anti-aliasing filter amplify the signal. I suspect it may be because the filter is not ideal or the algorithm introduce some kind of distorsion.

Below is the code:

[x,F]=audioread('example.wav');
Fs=F/24; %1837.5Hz
x_decimate=decimate(x,F/Fs,2,'iir');

Results:

max(x)
    1.000
min(x)
    -1.000
max(x_decimate)
    1.1108
min(x_decimate)
    -1.1078

What is wrong? Am I making a mistake?

Thank you

[Ameer Hamza]

What properties do you want in the downsampled signal? Should it be filtered during the downsampling process? You can also see downsample().

[Luciano Sebastián Rau]

I need to reduce the computation cost, so one technique is reduce the signal frequency.

The signal shold be filtered before the downsampling to be in order with the shannon-Nyquist criterion. The signal shold be filtered before the downsampling to be in order with the shannon-Nyquist criterion. The decimate() function is similar to do a filtered followed by a downsampling.

[Ameer Hamza]

What are the issues with values going beyond -1 and +1? I guess if the decimated signal looks similar to the original signal, then it should not be a problem. You can also consider rescaling then filtered so that it remains in the range of -1 to 1.

[Luciano Sebastián Rau]

In my algortihm I want to use a quantifier scheme in the future, so signal in the [-1,+1] range is a must. If I rescale the signal I need before to know how much, so it is no a good solution.

I only need to know why the decimated signal looks like it amplify the original signal. Maybe the decimate algorithm introduce some kind of distorsion.