| 摘要 |
<p>A conventional CELP speech codec (Figure 3) synthesizes a pitch interval in a sound by synthesizing a scaled innovation signal (44) -- typically, a random signal -- and adding it to a scaled pitch signal (56) derived from the synthesized speech (60) of the previous pitch interval. This invention (Figure 5) continues this practice when it is advantageous, but, at the onset of the sound or whenever else needed, replaces the scaled innovation signal (44) with a scaled spike signal (112). This is done since a spike (108) is sometimes more useful than an innovation signal (40) is, innovation signals (40) being by definition crafted to instead represent differences between adjacent pitch intervals within a sound rather than at the onset of a sound. <IMAGE></p> |
