| 摘要 |
A method and device for self-adaptive noise reduction. The method comprises: a first filter (210) is used to filter a signal received by a first microphone (110), a second filter (220) is used to filter a signal received by a second microphone (120), and a subtracter (300) is used to derive a noise-reduced signal by subtracting the filtered signals; in a noise segment, the noise-reduced signals are used to update respectively coefficients of the first filter (210) and of the second filter (220), thus the noise component that the signal filtered by the first filter (210) comprises and the noise component that the signal filtered by the second filter (220) comprises are brought to resemblance with each other; also, in a noisy speech segment, the coefficients of the first filter (210) and of the second filter (220) respectively are kept unchanged, while the coefficients updated in the preceding noise segment are used respectively by the first filter (210) and by the second filter (220) to filter the signals received by the first microphone (110) and by the second microphone (120). This solves the problem of poor noise reduction effects caused by the optimal solution for noise reduction being unreachable by using one FIR filter. |
