Implement Chromaprint wrapper algorithm

[dbogdanov]

Follow Chromaprint API.

Algorithm inputs:

• vector<StereoSample> audio
• Real sampleRate

Algorithm output:

• std::string fingerprint

Related to #474

Convert all values in input from Real to 16-bit int format to be able to use Chromaprint API.

[dbogdanov]

If possible, preferred input should be mono (vector<Real>) instead of stereo.

[palonso]

I have the wrapper working in Standard mode. There is a demostration retrieving the Acousticid id from the audio in this gist.
I have noticed that the algorithm is able to retrieve the id by just fingerprinting fragments instead of the full length of the track ( I used 30s segments in my example). So probably for the streaming mode it would be better to write an algorithm that retrieves a fingerprint each analysisSize seconds so it doesn't depend on previous knowledge about the input.

[dbogdanov]

That looks good. Correct, for the streaming mode we want to analyze on segments, but it would be great to still have an option to analyze on a complete input (if the analysisSize is set to 0).

[dbogdanov]

I guess the computed fingerprint should be concatenable, that is fingerprint(segment1+segment2) = fingerprint(segment1) + fingerprint(segment2). You can check if it is so in the gist.

[palonso]

There is a problem with that because in order to call the chromaprint api you have to specify the number of samples that it has to analyze. Standard mode is implemented as you say but I don't think it is possible in Streaming mode unless you specify the duration previously as an optional parameter.
I updated the gist with your proposed experiment.
Results suggest that it is possible to concatenate finguerprints and retrieve the AcouisticID key as far you include the proper song duration in the query.

[marv1nnnnn]

any update on this？

[dbogdanov]

@marv1nnnnn It is implemented in #601. To be merged soon.