Inactive Voice Signal Parameter Estimation for Comfort Noise Stability

The patent applies preliminary action by performing smooth processing on frequency spectrum coefficients before generating comfort noise. The encoder pre-processes the background noise characteristics during inactive voice phases, creating stabilized frequency spectrum parameters that are then transmitted to the decoder. This advance processing ensures that the comfort noise generated at the decoder side remains stable and free from bloop artifacts, resolving the reliability issue while maintaining the energy savings of DTX mode.

The patent applies parameter changes by transforming the frequency spectrum coefficients through smooth processing operations. Instead of transmitting raw background noise parameters that may be unstable, the encoder applies smoothing algorithms to modify the frequency spectrum parameters, creating more stable representations. These modified parameters are then transmitted over the channel, allowing the decoder to generate stable comfort noise even when the original background noise was large or unstable, thus maintaining reliability while conserving bandwidth.

The patent introduces an intermediary element in the form of smooth processing operations that act as a bridge between the raw frequency spectrum coefficients and the final comfort noise generation. The smooth processing serves as a mediator that stabilizes the parameters without requiring complex encoding schemes. By using relatively simple smoothing algorithms (such as exponential moving average or similar filters) on the frequency spectrum coefficients, the system achieves improved comfort noise stability while keeping the increase in encoding complexity minimal and manageable.