摘要 |
PURPOSE: An apparatus for quantizing a spectrum vector of an audio vector and a method for the same are provided to perform a weight vector quantization in consideration of the frequency masking effect and frequency resolution, thereby reducing the lowering of the quality of sound due to a quantization error. CONSTITUTION: An LP(Linear Prediction) analyzing member(10) predicts and analyzes linearly an input audio signal and then obtains spectrum envelope information and a residual signal. An FFT(Fast Fourier Transformation) member(20) converts the residual signal through the FFT transformation into a spectrum. An object vector generating member(30) generates an object vector from the spectrum applied from the FFT member. A perceptual weighted member(40) calculates a perceptual weighted function by using the LP coefficient calculated from the LP analyzing member, multiplies the error between the object vector and the codebook vector to the perceptual weighted function so that a weighted value in consideration of a frequency masking according to a band spectrum energy distribution is reflected. A bark weighted member(50) differentiates a bark frequency for representing the resolution of frequency so that a bark weighted function for reflecting the resolution of frequency relating to the sense of hearing in human is calculated. The bark weighted member multiplies the error to the bark weighted function so that the frequency resolution is reflected. An error minimization member(60) searches the codebook index for minimizing the error weighted at the bark weighted member from a codebook table. |