Earticle

To reproduce binaural sound at a listener’s ears using stereo loudspeakers in a portable device environment, it is important to design inverse filters that cancel out acoustical crosstalk signals. However, the direct and the crosstalk paths of a head- related transfer function are similar at certain frequencies, resulting in an excessive boost at those frequencies when inverse filters are designed. To mitigate this problem, a cross-talk cancellation filter design method is proposed which allows for the selective attenuation of unwanted peaks in the spectrum by constraining the magnitude of the difference between the direct and crosstalk paths. The performance of the proposed method is then evaluated by means of subjective source localization and objective tests. It is shown from the tests that the proposed method is able to provide binaural sound effects with very closely spaced stereo loudspeakers.

In super-resolution, various methods with Convolutional Neural Network(CNN) have recently been proposed. CNN based methods provide much higher image quality than conventional methods. Especially, VDSR outperforms other CNN based methods in terms of image quality. However, it requires a high computational complexity which prevents real-time processing. In this paper, the method to apply a deconvolution layer to VDSR is proposed to reduce computational complexity. Compared to original VDSR, the proposed method achieves the 4.46 times speed-up and its degradation in image quality is less than -0.1 dB which is negligible.