Nonnegative matrix factorization (NMF) is one of the most popular machine learning tools for speech enhancement (SE). However, there are two problems reducing the performance of the traditional NMFbased SE algorithms. One is related to the overlap-and-add operation used in the short time Fourier transform (STFT) based signal reconstruction, and the other is the Euclidean distance used commonly as an objective function; these methods can cause distortion in the SE process. In order to get over these shortcomings, we propose a novel SE joint framework which combines the discrete wavelet packet transform (DWPT) and the Itakura-Saito nonnegative matrix factorisation (ISNMF). In this approach, the speech signal was first split into a series of subband signals using the DWPT. Then, the ISNMF was used to enhance the speech for each subband signal. Finally, the inverse DWPT (IDWT) was utilised to reconstruct these enhanced speech subband signals. The experimental results show that the proposed joint framework effectively enhances the performance of speech enhancement and performs better in the unseen noise case compared to the traditional NMF methods.
In this paper, a modified sound quality evaluation (SQE) model is developed based on combination of an optimized artificial neural network (ANN) and the wavelet packet transform (WPT). The presented SQE model is a signal processing technique, which can be implemented in current microphones for predicting the sound quality. The proposed method extracts objective psychoacoustic metrics including loudness, sharpness, roughness, and tonality from sound samples, by using a special selection of multi-level nodes of the WPT combined with a trained ANN. The model is optimized using the particle swarm optimization (PSO) and the back propagation (BP) algorithms. The obtained results reveal that the proposed model shows the lowest mean square error and the highest correlation with human perception while it has the lowest computational cost compared to those of the other models and software.
The main objective of this paper is to produce an applications-oriented review covering infrared techniques and devices. At the beginning infrared systems fundamentals are presented with emphasis on thermal emission, scene radiation and contrast, cooling techniques, and optics. Special attention is focused on night vision and thermal imaging concepts. Next section concentrates shortly on selected infrared systems and is arranged in order to increase complexity; from image intensifier systems, thermal imaging systems, to space-based systems. In this section are also described active and passive smart weapon seekers. Finally, other important infrared techniques and devices are shortly described, among them being: non-contact thermometers, radiometers, LIDAR, and infrared gas sensors.
A traditional frequency analysis is not appropriate for observation of properties of non-stationary signals. This stems from the fact that the time resolution is not defined in the Fourier spectrum. Thus, there is a need for methods implementing joint time-frequency analysis (t/f) algorithms. Practical aspects of some representative methods of time-frequency analysis, including Short Time Fourier Transform, Gabor Transform, Wigner-Ville Transform and Cone-Shaped Transform are described in this paper. Unfortunately, there is no correlation between the width of the time-frequency window and its frequency content in the t/f analysis. This property is not valid in the case of a wavelet transform. A wavelet is a wave-like oscillation, which forms its own “wavelet window”. Compression of the wavelet narrows the window, and vice versa. Individual wavelet functions are well localized in time and simultaneously in scale (the equivalent of frequency). The wavelet analysis owes its effectiveness to the pyramid algorithm described by Mallat, which enables fast decomposition of a signal into wavelet components.