AudioEmbedding.h

Go to the documentation of this file.

#pragma once
 
namespace krotos
{
class AudioEmbedding
{
 
  public:
    AudioEmbedding();
 
    void initialise();
 
    std::vector<float> forward(const AudioSampleBuffer& buffer);
 
    std::vector<float> hz_to_mel(std::vector<float> freqs, bool htk = false);
 
    std::vector<float> mel_to_hz(std::vector<float> mels, bool htk = false);
 
    std::vector<float> mel_frequencies(float fmin, float fmax, int n_mels, bool htk = false);
 
    std::vector<std::vector<float>> mel(int sr, int n_fft, int n_mels, float fmin, float fmax, bool htk = false);
 
  private:
    void L2Normalise(std::vector<float>& x, float eps = 1e-5f);
 
    std::vector<float> applyMelFilterbank(const std::vector<float>& x);
 
    std::vector<float> applyAutoEncoder(const std::vector<float>& x);
 
    void writeMatrixToFile(const std::vector<std::vector<float>>& matrix, const std::string& filename);
 
    bool m_initialised;
    int m_fftSize;
    int m_hopSize;
    juce::dsp::FFT m_fft;
    juce::dsp::WindowingFunction<float> m_window;
    std::vector<std::vector<float>> m_mel;
 
    std::vector<std::vector<float>> m_weight = {
            {0.1183285f,  0.1893473f,  -0.0286719f, 0.0264190f,  0.2563457f,  0.0390199f,  0.2478790f,  0.0023484f,
             -0.2647046f, -0.0058535f, -0.0358872f, -0.0049668f, -0.2528938f, -0.3593057f, -0.0101721f, 0.0207894f,
             -0.0018700f, -0.0373632f, -0.3265873f, -0.0076127f, 0.3132295f,  0.0076612f,  -0.0280113f, -0.0444272f,
             -0.1723175f, -0.1415765f, -0.0550698f, -0.0235935f, 0.0207598f,  0.0717700f,  0.1228906f,  0.0681352f,
             -0.0029948f, -0.0291988f, 0.0164935f,  0.0638152f,  0.1206559f,  0.1614731f,  0.1714108f,  0.1438575f},
            {-0.2044093f, -0.0958435f, 0.1591669f,  0.1127563f,  -0.1829327f, -0.0261495f, -0.0692297f, -0.0076444f,
             -0.1753713f, -0.0086948f, -0.2606893f, -0.0011418f, 0.1213574f,  0.0206760f,  -0.0056649f, -0.3631761f,
             -0.0058598f, 0.1491575f,  -0.3030078f, -0.0005261f, 0.3438827f,  0.0266675f,  0.2152969f,  -0.0229872f,
             -0.1336009f, -0.0353414f, 0.1533862f,  0.3054215f,  0.1003580f,  -0.0969033f, -0.2173353f, -0.1318895f,
             -0.0001966f, 0.0937843f,  0.0670533f,  0.0350068f,  0.0365593f,  0.0486682f,  0.0322634f,  0.0380239f},
            {-0.0324681f, 0.3604097f,  0.2320765f,  -0.1200919f, 0.1243879f,  0.0169825f,  0.0191047f,  0.0064377f,
             0.2328371f,  0.0056822f,  -0.0142566f, 0.0053728f,  0.2740695f,  0.3054699f,  0.0122217f,  0.0910436f,
             -0.0029837f, -0.2393792f, -0.3604833f, -0.0116074f, -0.0441732f, 0.0145486f,  0.1619223f,  -0.0023186f,
             -0.1002556f, -0.1112942f, -0.1403950f, -0.2792160f, -0.1706164f, -0.1600267f, -0.2190082f, -0.1241049f,
             -0.0635069f, -0.0944659f, -0.0507775f, -0.0181387f, -0.0188597f, -0.0248787f, -0.0231442f, -0.0096551f},
            {-0.0568126f, 0.0039777f,  -0.0242762f, 0.0652586f,  0.2110767f,  0.0274380f,  -0.0389164f, -0.0059493f,
             -0.3283546f, -0.0061206f, -0.0439363f, -0.0081209f, -0.3111121f, 0.3638520f,  0.0172321f,  0.1167644f,
             -0.0029445f, -0.2031945f, 0.1368139f,  0.0084509f,  0.1455657f,  0.0047975f,  0.0066002f,  -0.0172313f,
             -0.0641237f, -0.0109537f, 0.0807085f,  0.1755439f,  0.0702213f,  -0.0292580f, -0.1117522f, -0.0761863f,
             -0.1240761f, -0.2945036f, -0.1867724f, -0.1091162f, -0.1381348f, -0.1887057f, -0.1823874f, -0.1375113f},
            {-0.0712475f, 0.1164975f,  -0.1592165f, -0.0468503f, 0.1823904f,  0.0235947f, -0.0005299f, -0.0061322f,
             -0.3679117f, -0.0087916f, -0.1065017f, 0.0069052f,  0.3553903f,  0.0421013f, -0.0006907f, -0.1037253f,
             -0.0003631f, 0.1638777f,  0.1558014f,  -0.0057055f, -0.3154457f, 0.0123267f, 0.2210963f,  0.0185530f,
             -0.0156829f, -0.0403669f, -0.0716844f, -0.1112856f, -0.0349633f, 0.0324615f, 0.0704213f,  0.0385223f,
             -0.0489561f, -0.1489882f, -0.0593635f, 0.0263974f,  0.0775172f,  0.1095617f, 0.1236699f,  0.1048189f},
            {0.1505479f,  0.0858491f,  -0.1504590f, -0.1528957f, -0.1403854f, -0.0235458f, -0.2887444f, -0.0050434f,
             0.0945120f,  0.0001072f,  -0.0692357f, -0.0063656f, -0.2482794f, -0.0585671f, 0.0005043f,  0.0280062f,
             -0.0067382f, -0.3248824f, -0.0238768f, -0.0084104f, -0.2341107f, 0.0285724f,  0.4004055f,  0.0341987f,
             0.0170751f,  0.0422470f,  0.1006104f,  0.1909740f,  0.0744385f,  -0.0191204f, -0.0611538f, -0.0442580f,
             -0.0435498f, -0.0690169f, 0.0021124f,  0.0699357f,  0.1302426f,  0.1843022f,  0.1867254f,  0.1634559f},
            {0.2622802f,  0.0747898f,  0.3714707f,  0.2570248f,  0.1412680f,  0.0127126f,  -0.2468822f, -0.0055563f,
             -0.0746559f, -0.0085515f, -0.3390156f, -0.0066121f, 0.0010087f,  -0.1027199f, -0.0109547f, -0.3698876f,
             -0.0116470f, -0.1509682f, 0.0146133f,  -0.0091458f, -0.3139470f, -0.0193163f, -0.1371502f, -0.0187797f,
             -0.0300480f, -0.0209954f, 0.0081653f,  0.0602221f,  0.0699154f,  0.1299464f,  0.1995567f,  0.1196174f,
             0.0730267f,  0.1058095f,  0.0195828f,  -0.0592565f, -0.1133265f, -0.1635979f, -0.1669433f, -0.1493805f},
            {-0.1151988f, -0.2073910f, -0.0786956f, 0.0412823f,  -0.0169307f, -0.0038584f, -0.1301079f, -0.0084627f,
             -0.2832127f, -0.0072423f, -0.1164781f, -0.0006129f, 0.0244106f,  -0.3093165f, -0.0159211f, -0.1414773f,
             -0.0128392f, -0.5125144f, 0.1679413f,  0.0157642f,  0.2505563f,  0.0122595f,  0.0666992f,  0.0635360f,
             0.1920382f,  0.0645521f,  -0.1973529f, -0.4753072f, -0.2420414f, -0.1009005f, -0.0660894f, -0.0249985f,
             0.0181937f,  0.0462444f,  0.0079372f,  -0.0182366f, -0.0393466f, -0.0511296f, -0.0505939f, -0.0509162f},
            {0.1418408f,  0.0163061f,  -0.2507642f, 0.2454780f,  0.1897282f,  0.0260308f,  0.0710254f,  0.0063145f,
             0.1383312f,  -0.0051422f, -0.4841823f, -0.0151193f, -0.2674182f, 0.0262099f,  0.0056830f,  0.2472064f,
             0.0101057f,  0.2991702f,  0.2853714f,  0.0163385f,  0.1941886f,  0.0502815f,  0.5026516f,  0.0348167f,
             -0.0568938f, -0.0663040f, -0.0765534f, -0.1684491f, -0.1147754f, -0.1222026f, -0.1589600f, -0.0796910f,
             0.0790012f,  0.2529074f,  0.1174290f,  -0.0028966f, -0.0532180f, -0.0738961f, -0.0943677f, -0.0819624f},
            {-0.0576470f, -0.1713927f, -0.2189870f, 0.5331324f,  0.1207931f,  0.0171014f,  0.1327211f,  0.0077573f,
             0.1837528f,  -0.0001906f, -0.1103368f, 0.0000165f,  0.1023141f,  0.0154868f,  0.0005956f,  0.0584098f,
             -0.0014845f, -0.2188623f, -0.2283842f, -0.0099906f, -0.0814616f, -0.0082530f, -0.0256755f, 0.0534589f,
             0.2403952f,  0.1815695f,  0.0552320f,  0.0365530f,  0.0119932f,  0.0058350f,  0.0056176f,  -0.0023686f,
             -0.0731922f, -0.1868461f, -0.0792575f, 0.0244835f,  0.0698310f,  0.1049344f,  0.1204482f,  0.1011122f},
            {0.0211549f,  0.2160409f,  -0.0257085f, 0.2807752f,  -0.1793017f, -0.0180747f, 0.3799123f,  0.0037635f,
             -0.1007019f, 0.0065484f,  0.4376744f,  0.0119324f,  0.1155654f,  -0.0804673f, -0.0081162f, -0.2667206f,
             -0.0141626f, -0.3772995f, 0.2177166f,  0.0068025f,  -0.1217531f, 0.0120425f,  0.1747560f,  -0.0014009f,
             -0.0577322f, -0.0107401f, 0.0807149f,  0.1557277f,  0.0534372f,  -0.0366322f, -0.0835269f, -0.0436425f,
             0.0866747f,  0.2498614f,  0.1096710f,  -0.0206487f, -0.0817196f, -0.1159209f, -0.1390758f, -0.1175552f},
            {-0.1557993f, -0.1381022f, 0.0918801f,  0.0980113f,  0.2412813f,  0.0248065f,  -0.3376282f, -0.0067565f,
             -0.1006937f, 0.0071258f,  0.3900249f,  0.0076727f,  -0.0371316f, -0.0410850f, 0.0018533f,  0.1872685f,
             0.0054611f,  0.0746087f,  -0.3026778f, -0.0159482f, -0.1079702f, 0.0229304f,  0.2912813f,  0.0143572f,
             -0.0497253f, -0.0478645f, -0.0232373f, -0.0190409f, 0.0079366f,  0.0497181f,  0.0943305f,  0.0626960f,
             0.1058429f,  0.2352961f,  0.1025504f,  -0.0116605f, -0.0534173f, -0.0790038f, -0.0922303f, -0.0884529f},
            {0.0600482f,  -0.1280766f, -0.2254330f, -0.1843846f, 0.3162802f, 0.0415772f,  -0.0398780f, 0.0030537f,
             0.0399624f,  0.0010333f,  -0.0088599f, 0.0042441f,  0.2194600f, 0.2194513f,  0.0030614f,  -0.2116873f,
             -0.0129591f, -0.3490874f, 0.1593306f,  0.0144684f,  0.3302665f, -0.0019648f, -0.1259521f, -0.0404725f,
             -0.1071220f, -0.0637591f, 0.0353417f,  0.1076920f,  0.0622471f, 0.0459982f,  0.0665063f,  0.0411561f,
             0.0879868f,  0.2211788f,  0.1297124f,  0.0531491f,  0.0528186f, 0.0655177f,  0.0576717f,  0.0431027f},
            {-0.1919733f, -0.0584059f, -0.1669482f, -0.1006971f, -0.1197154f, -0.0117195f, 0.2080063f,  0.0023273f,
             -0.0055921f, -0.0058544f, -0.2730973f, 0.0011942f,  0.2639267f,  -0.2841185f, -0.0008952f, 0.4410353f,
             0.0053654f,  -0.1691197f, -0.1730013f, -0.0133840f, -0.2086787f, -0.0123401f, -0.1239096f, -0.0959210f,
             -0.3431596f, -0.2217563f, 0.0274626f,  0.1782708f,  0.1034040f,  0.0543857f,  0.0386022f,  0.0205227f,
             0.0415918f,  0.0911896f,  0.0183219f,  -0.0626274f, -0.1186057f, -0.1746759f, -0.1861513f, -0.1512365f},
            {-0.1394873f, -0.3975483f, 0.1251428f,  -0.1387377f, 0.0806047f,  0.0196091f, 0.3981007f, 0.0149003f,
             0.2746806f,  0.0041046f,  -0.0753459f, -0.0109910f, -0.3234271f, 0.1851587f, 0.0029470f, -0.3338666f,
             -0.0130013f, -0.1404702f, -0.1215139f, -0.0170913f, -0.2611036f, 0.0147746f, 0.1958753f, -0.0198891f,
             -0.1780465f, -0.1820467f, -0.1518669f, -0.1904752f, -0.0400932f, 0.1066529f, 0.2078829f, 0.1211068f,
             0.0106500f,  -0.0427369f, -0.0172045f, 0.0119405f,  0.0414596f,  0.0539874f, 0.0642293f, 0.0500354f},
            {-0.0500911f, 0.0046412f, -0.2649742f, -0.1439137f, 0.3531356f,  0.0483050f,  0.0615251f,  0.0081148f,
             0.1698171f,  0.0069761f, 0.1044763f,  0.0012401f,  -0.1032799f, -0.2918410f, -0.0159013f, -0.3191282f,
             -0.0000891f, 0.1586788f, -0.0356244f, -0.0004457f, -0.2642753f, -0.0276544f, -0.1991710f, 0.0047946f,
             0.0992897f,  0.1062022f, 0.0704864f,  0.0179474f,  -0.0816622f, -0.2447657f, -0.3997843f, -0.2264303f,
             0.0049931f,  0.1480433f, 0.0715945f,  -0.0057449f, -0.0549305f, -0.0719114f, -0.0853416f, -0.0674315f},
            {-0.1100494f, 0.0360663f,  0.0767176f,  -0.1205767f, 0.1356053f,  0.0175366f,  -0.0061234f, 0.0069941f,
             0.3001839f,  0.0069099f,  0.0910342f,  0.0046054f,  0.0865339f,  -0.3466876f, -0.0145123f, -0.1996184f,
             -0.0065960f, -0.0180307f, 0.1514766f,  0.0064292f,  0.2911404f,  0.0382300f,  0.3457512f,  0.0212129f,
             -0.0319535f, -0.0120749f, 0.0469975f,  0.1519144f,  0.1062523f,  0.1070768f,  0.1162325f,  0.0533623f,
             -0.1315634f, -0.3953776f, -0.2362653f, -0.1060806f, -0.1052589f, -0.1405272f, -0.1264698f, -0.0930808f},
            {0.3935178f,  -0.3299477f, -0.0889016f, 0.0683443f,  -0.0365411f, -0.0066750f, -0.0660614f, -0.0024723f,
             -0.0388404f, 0.0040957f,  0.2137006f,  0.0098786f,  0.2172187f,  -0.0218917f, 0.0007969f,  0.0158916f,
             -0.0004570f, 0.0663592f,  -0.0219925f, -0.0004369f, 0.0257605f,  0.0092562f,  0.0570980f,  -0.0517055f,
             -0.2429105f, -0.1845663f, -0.0695577f, -0.0674032f, -0.0453340f, -0.0728290f, -0.1253623f, -0.0763488f,
             -0.0763951f, -0.1493674f, -0.0733747f, -0.0158430f, -0.0015317f, -0.0035468f, 0.0016920f,  0.0145657f},
            {0.2620155f,  -0.2097027f, 0.1695802f,  -0.3388280f, 0.0589687f,  0.0144293f,  0.3077674f,  0.0025493f,
             -0.2026729f, -0.0066413f, -0.0989198f, 0.0017403f,  0.1080217f,  -0.0253486f, 0.0009087f,  0.1589563f,
             0.0056932f,  -0.0447763f, -0.2211517f, -0.0073741f, -0.0373714f, 0.0089220f,  0.1779074f,  0.0968485f,
             0.3524145f,  0.2858000f,  0.1341488f,  0.1352768f,  0.0427499f,  -0.0133138f, -0.0378642f, -0.0197977f,
             0.0163499f,  0.0397683f,  -0.0070310f, -0.0442471f, -0.0911639f, -0.1156095f, -0.1242248f, -0.1102486f},
            {0.1160876f,  0.1843003f,  -0.5185968f, -0.0408614f, -0.1313047f, -0.0186611f, -0.0185707f, -0.0021685f,
             -0.0390545f, -0.0010064f, 0.0347686f,  -0.0000854f, -0.0163929f, 0.1319192f,  -0.0002808f, -0.1778814f,
             -0.0036161f, 0.0175012f,  -0.3065290f, -0.0130424f, 0.0771272f,  0.0136153f,  0.1394428f,  0.0409741f,
             0.1053845f,  0.0534246f,  -0.0335842f, -0.0515263f, 0.0288138f,  0.1546216f,  0.2580783f,  0.1529223f,
             0.0426801f,  -0.0121360f, -0.0651662f, -0.1116829f, -0.1786515f, -0.2464138f, -0.2486762f, -0.2183873f}};
};
} // namespace krotos