Robust bandwidth extension of narrowband speech signals

Speech signals contain frequencies between 100 Hz and 8 kHz (broadband speech). In telecommunication systems speech is transmitted in a frequency range between 300 Hz and 3.4 kHz (narrowband speech). This leads to a degradation of the intelligibility of the speech, since an important part of the frequency contents of fricatives like ‚f’ or ‚s‘ are in the upper band (4- 8 kHz).

This project aims at the implementation of an algorithm in MATLAB to estimate the upper band out of the narrow band without any additional information. Consequently, the speech sounds more natural and the intelligibility is being increased. The ability of the algorithm to be capable of working under real time conditions is an important design criteria too.