Suppression of Signal Peaks in Text-To-Speech Applications
Original Publication Date: 1988-Oct-01
Included in the Prior Art Database: 2005-Feb-15
This article describes the inclusion of an amplitude compression routine in the output stage of the synthesis filter when a DSP (Digital Signal Processor) is used to implement a speech synthesizer. This post-processing suppresses the unexpected signal peaks occasionally observed in text-to-speech applications. Background The intensity of resonance of an all-pole speech-synthesis filter is limited, and the stability of the filter is preserved by setting a threshold for the minimum distance between neighboring LSP (Line spectrum Pair) parameters in the frequency domain [*]. However, the output signal of such a stable synthesis filter can still have steep peaks when the interval and amplitude of each input excitation cause abnormal resonance of the filter.