WaveletsSample.java

package org.gennai.audio_analyzer.analysis;

import java.io.IOException;
import java.util.List;
import java.util.stream.Collectors;

import javax.sound.sampled.LineUnavailableException;
import javax.sound.sampled.UnsupportedAudioFileException;

import org.gennai.audio_analyzer.analysis.AudioTest.Audio;
import org.gennai.audio_analyzer.analysis.audio.AudioSamples;
import org.gennai.audio_analyzer.analysis.config.FingerprintConfig;
import org.gennai.audio_analyzer.analysis.config.QueryConfig;
import org.gennai.audio_analyzer.analysis.dao.IModelReference;
import org.gennai.audio_analyzer.analysis.dao.data.SoundData;
import org.gennai.audio_analyzer.analysis.data.Fingerprint;
import org.gennai.audio_analyzer.analysis.data.HashedFingerprint;
import org.gennai.audio_analyzer.analysis.hash.LocalitySensitiveHashingAlgorithm;
import org.gennai.audio_analyzer.analysis.inmemory.InMemoryModelService;
import org.gennai.audio_analyzer.analysis.inmemory.RAMStorage;
import org.gennai.audio_analyzer.analysis.query.QueryResult;
import org.gennai.audio_analyzer.analysis.query.ResultEntry;

public class WaveletsSample {

  private FingerprintConfig fingerprintConfig;
  private ModelService modelService;
  private FingerprintService fingerprintService;
  private LocalitySensitiveHashingAlgorithm lshAlgorithm;
  private QueryConfig queryConfig;
  private QueryFingerprintService queryFingerprintService;

  private WaveletsSample() {
    // 設定
    fingerprintConfig = new FingerprintConfig();
    fingerprintConfig.setTopWavelets(4000);
    /*
    fingerprintConfig.setNormalizeSignal(false);
    fingerprintConfig.getSpectrogramConfig().setWdftSize(2048);
    fingerprintConfig.getSpectrogramConfig().setOverlap(64);
    fingerprintConfig.getSpectrogramConfig().setFrequencyRange(318, 40000);
    fingerprintConfig.getSpectrogramConfig().setLogBins(32);
    fingerprintConfig.getSpectrogramConfig().setImageLength(128);
    */

    queryConfig = new QueryConfig();
    /*
    queryConfig.setThresholdVotes(5);
    queryConfig.setMaxSoundsToReturn(25);
    */

    // InMemory モデル
    RAMStorage storage = new RAMStorage(
        fingerprintConfig.getHashingConfig().getNumberOfLSHTables());
    modelService = new InMemoryModelService(storage);

    fingerprintService = new FingerprintService();
    lshAlgorithm = new LocalitySensitiveHashingAlgorithm();
    
    queryFingerprintService = new QueryFingerprintService();
  }

  private void insertHashData(String wavFile)
      throws UnsupportedAudioFileException, IOException, LineUnavailableException {
    // WAVEファイル読み込み
    Audio audio = AudioTest.readWave(wavFile);
    // audio.getData()   -> データ（double 配列）
    // audio.getOrigin() -> ファイルパス
    // audio.getFormat() -> AudioFormat（Android では MediaFormat に）

    AudioSamples audioSamples = new AudioSamples(audio.getData(), audio.getOrigin(),
        (int) audio.getFormat().getSampleRate());

    // Fingerprint 作成
    List<Fingerprint> fingerprints = fingerprintService.createFingerprints(audioSamples,
        fingerprintConfig);

    // Hashed fingerprint 作成
    List<HashedFingerprint> hashedFingerprints = fingerprints.parallelStream()
        .map(fingerprint -> lshAlgorithm.hash(fingerprint,
            fingerprintConfig.getHashingConfig().getNumberOfLSHTables(),
            fingerprintConfig.getHashingConfig().getNumberOfMinHashesPerTable(),
            fingerprintConfig.getClusters()))
        .collect(Collectors.toList());

    // メタデータをモデルに登録
    // 再生秒数は必須
    double length = (double) audio.getData().length / audio.getFormat().getSampleRate()
        / audio.getFormat().getChannels();
    // メタデータのリファレンスが返る
    IModelReference<?> soundReference = modelService.saveSound(new SoundData(length));

    // Hashed fingerprint をモデルに登録
    modelService.insertHashDataForSound(hashedFingerprints, soundReference);
  }

  private void query(String wavFile)
      throws UnsupportedAudioFileException, IOException, LineUnavailableException {
    // WAVEファイル読み込み
    Audio audio = AudioTest.readWave(wavFile);

    AudioSamples audioSamples = new AudioSamples(audio.getData(), audio.getOrigin(),
        (int) audio.getFormat().getSampleRate());

    // Fingerprint 作成
    List<Fingerprint> fingerprints = fingerprintService.createFingerprints(audioSamples,
        fingerprintConfig);

    // Fingerprint を Hashed fingerprint に変換
    List<HashedFingerprint> hashedFingerprints = fingerprints.parallelStream()
        .map(fingerprint -> lshAlgorithm.hash(fingerprint,
            fingerprintConfig.getHashingConfig().getNumberOfLSHTables(),
            fingerprintConfig.getHashingConfig().getNumberOfMinHashesPerTable(),
            fingerprintConfig.getClusters()))
        .collect(Collectors.toList());

    // モデルを検索
    QueryResult queryResult = queryFingerprintService.query(hashedFingerprints, queryConfig,
        modelService);
    // ヒットしたか？
    if (queryResult.containsMatches()) {
      // ベストマッチ
      ResultEntry resultEntry = queryResult.bestMatch();
      System.out.printf("best id: %d, score: %d\n",
          resultEntry.getSound().getSoundReference().getId(),
          resultEntry.getHammingSimilaritySum());

      // マッチリスト
      queryResult.getResultEntries().forEach(entry -> {
        System.out.printf("id: %d, score: %d\n",
            entry.getSound().getSoundReference().getId(), entry.getHammingSimilaritySum());
      });
    } else {
      System.out.println("no match");
    }
  }
  
  public static void main(String[] args)
      throws UnsupportedAudioFileException, IOException, LineUnavailableException {
    WaveletsSample waveletsSample = new WaveletsSample();
    waveletsSample.insertHashData("../samples4/nom1.wav");
    waveletsSample.query("../samples4/nom1.wav");
  }
}