This paper introduces a new noise generation algorithm for vocoder-based speech waveform generation. White noise is generally used for generating an aperiodic component. Since short-term white noise includes a zero-frequency component (ZFC) and inaudible components below 20 Hz, they are reduced in advance when synthesizing. We propose a new noise generation algorithm based on that for velvet noise to overcome the problem. The objective evaluation demonstrated that the proposed algorithm can reduce the unwanted components.

Noise characteristics of GaAs metal-semiconductor field effect transistors (GaAs-MESFETs) with scaled-down dimensions are analyzed and modeled using a physics-based circuit simulator employing the Monte Carlo (MC) particle technique. The microscopic dynamics of electrons is also analyzed to investigate the mechanism of noise generation in a channel. Noise spectral densities of GaAs-MESFETs with two different geometries are estimated by evaluating fluctuations in instantaneous terminal currents. Then, minimum noise figures, F min, and noise figure circles are estimated using the noise spectral densities and Y-parameters. Because of an increase in y21 and suppression of an increase of noise spectral density, the device with an n+-region extending to below the drain-side edge of the gate contact exhibits a smaller noise figure. Suppression of the electron velocity fluctuation caused by electron transitions to higher valleys in a high electric field region is responsible for the noise suppression.

A flexible noise generation algorithm using DCT is proposed. The proposed method outperforms the conventional methods when a noise model requires complicated PSD (Power Spectral Density) specifications. Also, it is shown that the proposed system can be used for the test of VDSL (Very high-speed Digital Subscriber Line).