The test audio used in this project (Am I Blue.wav) is Kevin Conroy's recording of Am I Blue, available on his website. I wrote my own simplified MATLAB code to compute the discrete forward and. While mean-squared-error was used as the error metric, a human listener will also often prefer the DWT-compressed signal, possibly since our ears rely partially on a DWT (see the paper, which gives a reference for this claim). called the Haar wavelet, thus there would be small finite amount of expansion. The results, shown in the paper, are that in some cases DWT can get better compression for the same mean-squared-error in the decompressed signal than DCT. The code is designed to compute Daubechies wavelet coefficients of arbitrary order, apply a DWT to an audio signal, apply truncated wavelet approximation to compress the signal with minimal losses, and decompress/transform to recover an audio signal.įurthermore, it compares truncated wavelet approximation (using several sets of Daubechies coefficients) to truncated approximation using discrete cosine series. The paper (stored in this repository as a PDF) reviews the discrete wavelet transform (DWT) and explains this code. The wavedec function from the MATLAB wavelet toolbox is responsible for implementing the db2 transform (four filter coefficients) with their respective levels of decomposition. This is my final project for the MIT course 18.335, Introduction to Numerical Methods, written in Julia. In this section the wavelet transform is applied to the 1 sec segments obtained in the previous section. Audio Compression via Discrete Wavelet Transform
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |