We address the speaker independent automatic recognition of spontaneous speech in highly variable noise by applying semisupervised sparse non-negative matrix factorization (NMF) for speech enhancement coupled with our recently proposed frontend utilizing bottleneck (BN) features generated by a bidirectional Long Short-Term Memory (BLSTM) recurrent neural network. In our evaluation, we unite the...

The gamut of a colour space is defined by a number of extreme points. The best inks to achieve an accurate spectral reproduction of a given target are those which span the target’s spectral gamut. Using a modified non-negative matrix factorization (NMF) algorithm we derive m colorants and their spectral curves such that they are the extreme points of the targets gamut. Using the spectral Neugeb...

This paper describes a face detection approach via learning local features. The key idea is that local features, being manifested by a collection of pixels in a local region, are learnt from the training set instead of arbitrarily defined. The learning procedure consists of two steps. First, a modified version of NMF (Non-negative Matrix Factorization), namely local NMF (LNMF), is applied to ge...

This study presents an automatic glottal inverse filtering (GIF) technique based on separating the effect of the glottal main excitation from the impulse response of the vocal tract. The proposed method is based on a non-negative matrix factorization (NMF) based decomposition of an ultra short-term spectrogram of the analyzed signal. Unlike other state-of-theart GIF techniques, the proposed met...

Nonnegative matrix factorization (NMF) aims at factoring a data matrix into low-rank latent factor matrices with nonnegativity constraints on (one or both of) the factors. Specifically, given a data matrix X ∈ RM×N and a target rank R, NMF seeks a factorization model X ≈WH>, W ∈ RM×R, H ∈ RN×R, to ‘explain’ the data matrix X, where W ≥ 0 and/or H ≥ 0 and R ≤ min{M,N}. At first glance, NMF is no...

In this paper, we investigate the performance of classifierbased non-negative matrix factorization (NMF) methods for detecting overlapping acoustic events. We provide evidence that the performance of classifier-based NMF systems deteriorates significantly in overlapped scenarios in case mixed observations are unavailable during training. To this end, we propose a K-means based method for artifi...

Non-negative matrix factorization (NMF) has previously been shown to be a useful decomposition for multivariate data. Two different multiplicative algorithms for NMF are analyzed. They differ only slightly in the multiplicative factor used in the update rules. One algorithm can be shown to minimize the conventional least squares error while the other minimizes the generalized Kullback-Leibler d...

A novel Non-negative Matrix Factorization (NMF) based subband decomposition in frequency-spatial domain for acoustic source localization using a microphone array. The proposed method decomposes source and noise subband and emphasizes source dominant frequency bins for more accurate source representation. By employing NMF, we extract Delay Basis Vectors (DBV) and their subband information in fre...

Changing illumination causes severe problems for face recognition in uncontrolled environments. It might be helpful for illumination invariant face recognition if information about the illumination can be recovered from the given face image. In this paper an illumination classification method based on Non-negative Matrix Factorization(NMF) is proposed. The traditional NMF approach together with...

Image hashing is an efficient technique for indexing and can be effectively used in image retrieval .This paper uses image hashing with a ring partition and a non-negative matrix factorization(NMF),which provides rotation robustness and good discriminative capability respectively. The aim of ring partition in image hashing is to construct a rotation invariant secondary image for dimensionality ...