The “dsp.SampleRateConverter” function in MATLAB requires both the “InterpolationFactor” and “DecimationFactor” to be integer-valued. The error occurs because the expected `InterpolationFactor` is not integer-valued, likely due to a floating-point value derived from a sample rate ratio being passed into it.
Rounding the interpolation factor to an integer is a good workaround. For deep learning training, such as in the Modulation Classification example, this typically has minimal impact, especially if the rounded factor still produces a resampling ratio close to the original. Deep learning models are generally resilient to slight timing or sampling variations in the input.
However, rounding can slightly alter the effective sample rate, which may affect signal fidelity in edge cases—particularly for classes like ‘B-FM’ that are sensitive to frequency content. To ensure the resampling ratio is accurately maintained, consider enhancing the approach by using:
[L, M] = rat(desiredRatio, 1e-6); % Rational approximation to keep integers
This method keeps both factors as integers while closely matching the original ratio.
Refer to the following MathWorks documentation for more information:
Hope it works!
