摘要 |
PURPOSE:To improve the converging characteristic even to a reception side input signal intensive in correlativity by providing a means reducing the correlativity based on a self-recurring model. CONSTITUTION:When a signal of a predetermined time length is inputted, a forecast coefficient calculator 22 regards it as an output signal of the self-recurring model, a linear forecasting coefficient ai is obtained to apply stability discrimination when a recursive filter is constituted, the value of the linear forecasting coefficient is limited to ensure the stability of the recursive filter. In case of a linear forecast coefficient aM up to the M-th order, e.g., a linear forecast coefficient a2, when values a1, a2 exist in a triangle region, the recursive filter is activated stable. When the values a1, a2 exist at the outside of the triangle region conversely, the values a1, a2 are switched forcibly in values within the region. |