Time-Frequency transformation and spectral representations of audio signals are commonly used in various machine learning applications. Typically the Mel-Spectrogram is used to create the input features to the network justified by the Mel scale’s human auditory system basis. In this paper, we compare several spectral features in a gender detection speech model comparing their performance and showing that the Mel-Spectrogram is not always the best choice for input features.
(See document for exact affiliation information.)
https://aes2.org/publications/elibrary-page/?id=21963
(470KB)
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member Join the AES. If you need to check your member status, login to the Member Portal.
