We propose deep neural network (DNN)-based speech enhancement that reduces musical noise and achieves better auditory impressions. The musical noise is an artifact generated by nonlinear signal processing and negatively affects the auditory impressions. We aim to develop musical-noise-free speech enhancement methods that suppress the musical noise generation and produce perceptually-comfortable enhanced speech. DNN-based speech enhancement using a soft mask achieves high noise reduction but generates musical noise in non-speech regions. Therefore, first, we define kurtosis matching for DNN-based low-musical-noise speech enhancement. Kurtosis is the fourth-order moment and is known to correlate with the amount of musical noise. The kurtosis matching is a penalty term of the DNN training and works to reduce the amount of musical noise. We further extend this scheme to standardized-moment matching. The extended scheme involves using moments whose orders are higher than kurtosis and generalizes the conventional musical-noise-free method based on kurtosis matching. We formulate standardized-moment matching and explore how effectively the higher-order moments reduce the amount of musical noise. Experimental evaluation results 1) demonstrate that kurtosis matching can reduce musical noise without negatively affecting noise suppression and 2) newly reveal that the sixth-moment matching also achieves low-musical-noise speech enhancement as well as kurtosis matching.

In a speech enhancement system for impact noise, it is important for any impact noise activity to be detected. However, because impact noise occurs suddenly, it is not always easy to detect. We propose a method for impact noise activity detection based on the kurtosis of an instantaneous power spectrum. The continuous duration of a generalized impact noise is shorter than that of speech, and the power of such impact noise varies dramatically. Consequently, the distribution of the instantaneous power spectrum of impact noise is different from that of speech. The proposed detection takes advantage of kurtosis, which depends on the sharpness and skirt of the distribution. Simulation results show that the proposed noise activity detection improves the performance of the speech enhancement system.

This letter investigates the circularity of fractional Fourier transform (FRFT) coefficients containing noise only, and proves that all coefficients coming from white Gaussian noise are circular via the discrete FRFT. In order to use the spectrum kurtosis (SK) as a Gaussian test to check if linear frequency modulation (LFM) signals are present in a set of FRFT points, the effect of the noncircularity of Gaussian variables upon the SK of FRFT coefficients is studied. The SK of the α th-order FRFT coefficients for LFM signals embedded in a white Gaussian noise is also derived in this letter. Finally the signal detection algorithm based on FRFT and SK is proposed. The effectiveness and robustness of this algorithm are evaluated via simulations under lower SNR and weaker components.

In this letter, a new approach to segment depth-of-field (DoF) images is proposed. The methodology is based on a two-stage model of visual neuron. The first stage is a retinal filtering by means of luminance normalizing non-linearity. The second stage is a V1-like filtering using filters estimated by independent component analysis (ICA). Segmented image is generated by the response activity of the neuron measured in terms of kurtosis. Results demonstrate that the model can discriminate image parts in different levels of depth-of-field. Comparison with other methodologies and limitations of the proposed methodology are also presented.

In this paper, we introduce a generalized minimum mean-square error short-time spectral amplitude estimator with a new prior estimation of the speech probability density function based on moment-cumulant transformation. From the objective and subjective evaluation experiments, we show the improved noise reduction performance of the proposed method.

In this paper, we propose a robust room inverse filtering algorithm for speech dereverberation based on a kurtosis maximization. The proposed algorithm utilizes a new normalized kurtosis function that nonlinearly maps the input kurtosis onto a finite range from zero to one, which results in a kurtosis warping. Due to the kurtosis warping, the proposed algorithm provides more stable convergence and, in turn, better performance than the conventional algorithm. Experimental results are presented to confirm the robustness of the proposed algorithm.

The interference imposed on conventional narrow-band systems by impulse radio UWB (IR-UWB) signals is examined by simulations. The Dirac delta function is employed to model the IR-UWB signal to reduce simulation costs. The simulation results show that the statistical characteristics of this interference deviate from Gaussian noise when the frequency band of the narrow-band system includes a half multiple of the data symbol rate of the IR-UWB system. In the case of pulse-position-modulation UWB signals and biorthogonal-coded bipolar-modulation UWB signals, the performance degradation of the narrow-band system depends on the number of pulse positions and the number of orthogonal codes, respectively.