aubio MFCC extractor swift

aubioHelper.swift

import Foundation
import aubio

class AubioHelper {
    
    private let audioCapacity = 24000
    private let hop_size: uint_t = 1024
    private let win_size: uint_t = 4096
    private let buf_size: uint_t = 4096
    private let n_filters: uint_t = 40
    private let n_coeffs: uint_t = 13
    
    func getMFCC(data: [Float32]) -> [[Float32]] {
        let sample_rate: uint_t = 16000        
        
        let samples = new_fvec(buf_size);
        var mfcc = UnsafeMutablePointer<Float32>.allocate(capacity: audioCapacity)
        let mfcc_spec = new_aubio_mfcc(buf_size, n_filters, n_coeffs, sample_rate)
        let samples_cvec = new_cvec(win_size)
        let pvoc = new_aubio_pvoc(win_size, hop_size)
        let out = new_fvec(n_coeffs)

        var sampleCount: UInt32 = 0
        var total_frames : uint_t = 0
        var mfcc_array: [[Float32]] = []
        
        for i in 0..<data.count {
            fvec_set_sample(samples, data[i], sampleCount)
            sampleCount += 1
                                
            if sampleCount >= hop_size {
                                            
                //TODO: Already tried silence threshold but didn't work
                                                    
                // calculate MFCCs from audio values
                aubio_pvoc_do(pvoc, samples, samples_cvec)
                aubio_mfcc_do(mfcc_spec, samples_cvec, out)
                mfcc = fvec_get_data(out)
                
                mfcc_array.append(Array(UnsafeBufferPointer(start: mfcc, count: Int(n_coeffs))))
                
                total_frames += sampleCount
                sampleCount = 0
            }
        }        
        
        // destroy used vectors
        del_fvec(samples)
        del_aubio_mfcc(mfcc_spec)
        del_cvec(samples_cvec)
        del_aubio_pvoc(pvoc)
        del_fvec(out)
                
        return mfcc_array
    }
}