- Take a look at this survey paper: Comprehensive Survey on Distance/Similarity Measures between Probability Density Functions
- I once was working on quantifying the similarity between spiky signals (i.e., signals with fast decay), the normalized mutual information was the best measure for me.
- Experiment with several measures, until you find the one satisfactory for your application.
Comparing Audio Impulse Responses
1 view (last 30 days)
This is not specifically a matlab question. However, I would ask the question here anyway.
I generate impulse responses in shoebox room using a toolbox based on the image method. With every parameter same except for the reflection order, I want to compare how similar the impulse responses are.
As the impulse responses are a vector, can cosine similarity be used? Is there any other method to check how similar impulse responses are?
Yazan on 11 Aug 2021
More generally, I believe you're looking for a quantitative measure of similarity between functions. This is a large domain of research. The best measure will always depend on the problem and signals at hand, so there is no simple answer. I can recommend the following: