Moving-Horizon Optimal Quantizer for Audio Signals
By analyzing the quantization of audio signals as a deterministic finite-set constrained quadratic optimization problem, a new scheme, called moving-horizon optimal quantizer (MHOQ), is developed. The MHOQ includes a model of the ear's sensitivity to low-level noise power and minimizes directly the perceived error over a finite prediction horizon. Feedback is incorporated by means of the moving-horizon principle. With a prediction horizon equal to 1, the MHOQ reduces to the psychoacoustically optimal noise-shaping quantizer, widely used in practical applications. Larger prediction horizons outperform the noise shaper at the expense of only a small increase in computational complexity.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is temporarily free for AES members.