Recently, I am using CO2 signal from Capnobase dataset to estimate the respiratory rate.

What I have done is to choose a 30-second data (sampling rate = 300Hz, 9000 samples), use Fourier Bessel Series Expansion to get the magnitude spectrum, find the frequency corresponding to the highest peak, and multiply the frequency with 60.

However, the result is not precise because the interval between two respiratory rate is 1.

The paper I referred to is " FBSE-EWT-Based Approach for the Determination of Respiratory Rate From PPG Signals ", and the points in the paper look much close to each other.

Are there any possible parameters to influence the precision? (Maybe from FBSE coefficient formula)

Or does the normalization of frequency in spectrum to [0, pi] lead to this problem? (I didn't normalize the frequency from 150 to pi)