摘要
A simple ultraviolet (UV) decision making method using sub bands can greatly reduce the computational complexity and the memory requirements in the MBE LPC algorithm and can reduce the difficulties in DSP real time fixed point implementation and integrated circuit chips. The frequency waveform of each sub band of the linear prediction coding (LPC) residue is used to calculate the autocorrelation function. Its periodicity is analyzed to determine if it is voice or unvoice. The principles of this UV decision are compatible with that of the full band speech UV decision. A subjective listening test showed that the speech synthesized by this method is not inferior to that of the standard multi band exitation (MBE) algorithm.
A simple ultraviolet (UV) decision making method using sub bands can greatly reduce the computational complexity and the memory requirements in the MBE LPC algorithm and can reduce the difficulties in DSP real time fixed point implementation and integrated circuit chips. The frequency waveform of each sub band of the linear prediction coding (LPC) residue is used to calculate the autocorrelation function. Its periodicity is analyzed to determine if it is voice or unvoice. The principles of this UV decision are compatible with that of the full band speech UV decision. A subjective listening test showed that the speech synthesized by this method is not inferior to that of the standard multi band exitation (MBE) algorithm.
基金
the National Natural Science Foundationof China(No.6 96 36 0 30 )
