bhlangonijr / classifier

Experimental project for classifying voice command intention

Audio Classifier Using Statistical Learning

Experimental project for voice command intention detection based on statistical learning Java FFT and audio analysing took from audio-analysis

Testing

Execute the unit tests. More to come...

    val featureNames = mutableMapOf<String, Int>()
    for (i in 1..30) {
        featureNames["$i"] = i
    }
    
    // audio samples for training: 0.0 = YES 1.0 = NO
    val yes2 = audioToFs(0, "src/test/resources/yes2.wav", featureNames, 0.0)
    val yes3 = audioToFs(1, "src/test/resources/yes3.wav", featureNames, 0.0)
    val yes4 = audioToFs(2, "src/test/resources/yes4.wav", featureNames, 0.0)
    val no2 = audioToFs(3, "src/test/resources/no2.wav", featureNames, 1.0)
    val no3 = audioToFs(4, "src/test/resources/no3.wav", featureNames, 1.0)
    val no4 = audioToFs(5, "src/test/resources/no4.wav", featureNames, 1.0)
    
    
    val data = DataSet(arrayListOf(yes2, yes3, yes4, no2, no3, no4), featureNames.keys.toList())
    //println(data)
    
    val nb = NaiveBayes()
    
    // not known samples
    val yes1 = audioToFs(6, "src/test/resources/yes1.wav", featureNames, -1.0)
    val no1 = audioToFs(7, "src/test/resources/no1.wav", featureNames, -1.0)
    
    val stats = nb.train(data)
    
    //println(stats)
    
    val classification1 = nb.classify(yes1, stats)
    val classification2 = nb.classify(no1, stats)
    
    println(classification1.predictions)
    println(classification1.predict())
    
    println(classification2.predictions)
    println(classification2.predict())
    
    assertEquals(0.0, classification1.predict())
    assertEquals(1.0, classification2.predict())