abstract
- This work presents the use of the k-nearest neighbors (kNNs) technique to achieve the classification of selected sounds through the use of audio analysis. By implementing some audio spectral analysis, it is possible to establish a unique trace of the sound from each audio file. These traces are used to make computers learn to differentiate between various types of sounds in acoustic scenes by implementing the kNN machine learning technique. Adding the principal component analysis (PCA) technique for processing the data before including it in the classification model is expected to increase the accuracy of the model compared to the case when PCA is not introduced. The results show that the PCA technique improves the kNN model by analyzing the data distribution and reducing the dimensions of the data set keeping the most representative information of the original data. © 2024, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.