How to correlate a sample audio to a train dataset

I have a trained dataset F(70,5) .. (70row 5columns) . How do I correlate sample audio to this whole trained class and see the relationship?

I want to corellate their spectrum and cepstrum.

[Walter Roberson]

Is that 70 features and 5 samples? 5 features and 70 samples? 5 samples of raw audio at 70 samples each?

[Kamil Kacer]

In the train class there is 5 samples and each has 70 features

[Walter Roberson]

For each audio sample, extract its features, and calculate the euclidean distance between those features and the 70-dimensional points (note: each component should be normalized.) The one out of the five that is nearest is the most similar.