Time needed for calculation

[TimAmadeoSobania]

Hi, I am currently trying to use tsfresh for something it says it is not suitable for. . I have an algorithm that separates cycles in streaming data. Now as I separate those cycles they can be seen as historical and therefore I hope to accomplish a comparison of the cycles with tsfresh.
My problem is that the feature extraction I do with about 22 features (see below) is taking at least 3.5seconds even with only one cycle. This means that I could only use it for data where one cycle is taking a longer timespan than 3.5 seconds.
Now, is there something I could do to speed the extraction up? And the other question is what all is contibuting to those 3.5s? As it still takes ~3.5 seconds even if I only select one feature it probably isnt the mathematical calculation.

My Code:

fc_parameters = {
"length": None,
"abs_energy": None,
"benford_correlation":None,
"count_above_mean": None,
"first_location_of_maximum": None,
"first_location_of_minimum": None,
"last_location_of_maximum": None,
"last_location_of_minimum": None,
"has_duplicate_max": None,
"has_duplicate_min": None,
"maximum": None,
"minimum": None,
"mean": None,
"mean_change": None,
"median": None,
"root_mean_square": None,
"sample_entropy": None,
"standard_deviation": None,
"sum_values": None,
"variance": None,
"variance_larger_than_standard_deviation": None,
"variation_coefficient": None,
}

extracted_features = tsf.extract_features(df, column_id="cycle", default_fc_parameters= fc_parameters)

The dataframe consists of 141 float values.