Analisis Perbandingan Kompresi Suara Menggunakan Principal Component Analysis dan Transformasi Wavelet

One of the requirements faced as a result of information technology development is memory and transmission efficiency. This requirement can be overcome with data compression. Compression is a method to obtain compact data with a smaller size but still maintaining similarity to the original data. Principal Component Analysis (PCA) is an algorithm in machine learning that is used to reduce dimensions. Dimensional reduction is a process of transforming high-dimensional data into new subspaces with lower dimensions. The goal is to use some principal components to represents the original data. Wavelet transformation represents a signal into a set of basic functions through filter analysis. Wavelets concentrate information into coefficients of approximation and coefficients of detail. Wavelet transform produces a lot of zero or close to zero coefficients that can be neglected so it can reduce storage space. In this research, we will propose the implementation of PCA and Wavelet for digital audio compression. The audio was performed with the .wav format. The compressed audio will be evaluated based on Peak Signal to Noise Ratio (PSNR) and Mean Square Error (MSE). The mean PSNR obtained when using a wavelet is 47.61601 dB  with an average MSE of 3.76 x 10-5. Meanwhile, when using PCA, the PSNR average was 57.3962772 dB and the average MSE obtained was 4.59 x 10-5. Four out of five compressed audio had a larger PSNR and smaller MSE when using PCA. Thus, the Principal Component Analysis algorithm can be better used for audio compression than the level 1 of Symlet Wavelet Transformation.