Figure 10From: Single-channel noise reduction using unified joint diagonalization and optimal filteringSpectra of desired signal, noisy signal and two reconstructions with different choices of P′. (a) Spectrum of a part of a speech signal from the Keele database. (b) Speech signal from (a) contaminated with babble noise to give an average input SNR of 10 dB. (c) Reconstructed speech signal using P′=10. (d) Reconstructed speech signal using P′=70.Back to article page