In the development process of noise-reduction algorithms, an objective machine-driven intelligibility measure which shows high correlation with speech intelligibility is of great interest. Besides reducing time and costs compared to real listening experiments, an objective intelligibility measure could also help provide answers on how to improve the intelligibility of noisy unprocessed speech. In this paper, a short-time objective intelligibility measure (STOI) is presented, which shows high correlation with the intelligibility of noisy and time-frequency weighted noisy speech (e.g., resulting from noise reduction) of three different listening experiments. In general, STOI showed better correlation with speech intelligibility compared to five other reference objective intelligibility models. In contrast to other conventional intelligibility models which tend to rely on global statistics across entire sentences, STOI is based on shorter time segments (386 ms). Experiments indeed show that it is beneficial to take segment lengths of this order into account. In addition, a free Matlab implementation is provided.

An Algorithm for Intelligibility Prediction of Time–Frequency Weighted Noisy Speech

This paper describes a public-domain database of high-spatial-resolution head-related transfer functions measured at the UC Davis CIPIC Interface Laboratory and the methods used to collect the data.. Release 1.0 (see http://interface.cipic.ucdavis.edu) includes head-related impulse responses for 45 subjects at 25 different azimuths and 50 different elevations (1250 directions) at approximately 5/spl deg/ angular increments. In addition, the database contains anthropometric measurements for each subject. Statistics of anthropometric parameters and correlations between anthropometry and some temporal and spectral features of the HRTFs are reported.

https://escholarship.org/content/qt3d10j9jw/qt3d10j9jw.pdf?t=ptt2wh

The CIPIC HRTF database

Speech separation is the task of separating target speech from background interference. Traditionally, speech separation is studied as a signal processing problem. A more recent approach formulates speech separation as a supervised learning problem, where the discriminative patterns of speech, speakers, and background noise are learned from training data. Over the past decade, many supervised separation algorithms have been put forward. In particular, the recent introduction of deep learning to supervised speech separation has dramatically accelerated progress and boosted separation performance. This paper provides a comprehensive overview of the research on deep learning based supervised speech separation in the last several years. We first introduce the background of speech separation and the formulation of supervised separation. Then, we discuss three main components of supervised separation: learning machines, training targets, and acoustic features. Much of the overview is on separation algorithms where we review monaural methods, including speech enhancement (speech-nonspeech separation), speaker separation (multitalker separation), and speech dereverberation, as well as multimicrophone techniques. The important issue of generalization, unique to supervised learning, is discussed. This overview provides a historical perspective on how advances are made. In addition, we discuss a number of conceptual issues, including what constitutes the target source.

Supervised Speech Separation Based on Deep Learning: An Overview

There has been a great deal of material in the area of discrete-time transforms that has been published in recent years. This book does an excellent job of presenting important aspects of such material in a clear manner. The book has 11 chapters and a very useful appendix. Seven of these chapters are essentially devoted to the Fourier series/transform, discrete Fourier transform, fast Fourier transform (FFT), and applications of the FFT in the area of spectral estimation. Chapters 8 through 10 deal with many other discrete-time transforms and algorithms to compute them. Of these transforms, the KarhunenLoeve, the discrete cosine, and the Walsh-Hadamard transform are perhaps the most well-known. A lucid discussion of number theoretic transforms i5 presented in Chapter 11. This reviewer feels that the authors have done a fine job of compiling the pertinent material and presenting it in a concise and clear manner. There are a number of problems at the end of each chapter, an appreciable number of which are challenging. The authors have included a comprehensive set of references at the end of the book. In brief, this book is a valuable contribution to the general areas of signal processing and communications. It can be used for a graduate level course in perhaps two ways. One would be to cover the first seven chapters in great detail. The other would be to cover the whole book by focussing on different topics in a selective manner. This book  by  Elliott  and Rao  is extremely  useful to researchers/engineers who are working  in the areas of signal processing and communications. It i s also an excellent reference book, and hence a valuable addition to one’s library

http://ieeexplore.ieee.org/iel5/5/31347/01457444.pdf

Multirate digital signal processing

A computational model of auditory analysis is described that is inspired by psychoacoustical and neurophysiological findings in early and central stages of the auditory system The model provides a unified multiresolution representation of the spectral and temporal features likely critical in the perception of sound Simplified, more specifically tailored versions of this model have already been validated by successful application in the assessment of speech intelligibility [Elhilali et al, Speech Commun 41(2-3), 331-348 (2003); Chi et al, J Acoust Soc Am 106, 2719-2732 (1999)] and in explaining the perception of monaural phase sensitivity [R Carlyon and S Shamma, J Acoust Soc Am 114, 333-348 (2003)] Here we provide a more complete mathematical formulation of the model, illustrating how complex signals are transformed through various stages of the model, and relating it to comparable existing models of auditory processing Furthermore, we outline several reconstruction algorithms to resynthesize the sound from the model output so as to evaluate the fidelity of the representation and contribution of different features and cues to the sound percept

https://ir.nctu.edu.tw:443/bitstream/11536/13455/1/000231210500036.pdf

Multiresolution spectrotemporal analysis of complex sounds

Monaural spectral features due to pinna diffraction are the primary cues for elevation. Because these features appear above 3 kHz where the wavelength becomes comparable to pinna size, it is generally believed that accurate elevation estimation requires wideband sources. However, psychoacoustic tests show that subjects can estimate elevation for low-frequency sources. In the experiments reported, random noise bursts low-pass filtered to 3 kHz were processed with individualized head-related transfer functions (HRTFs), and six subjects were asked to report the elevation angle around four cones of confusion. The accuracy in estimating elevation was degraded when compared to a baseline test with wideband stimuli. The reduction in performance was a function of azimuth and was highest in the median plane. However, when the source was located away from the median plane, subjects were able to estimate elevation, often with surprisingly good accuracy. Analysis of the HRTFs reveals the existence of elevation-dependent features at low frequencies. The physical origin of the low-frequency features is attributed primarily to head diffraction and torso reflections. It is shown that simple geometrical approximations and models of the head and torso explain these low-frequency features and the corresponding elevations cues.

/pdf/elevation-localization-and-head-related-transfer-function-3ixmbm0zse.pdf

Elevation localization and head-related transfer function analysis at low frequencies.

Systems and methods for providing single microphone noise suppression fallback are provided. In exemplary embodiments, primary and secondary acoustic signals are received. A single microphone noise estimate may be generated based on the primary acoustic signal, while a dual microphone noise estimate may be generated based on the primary and secondary acoustic signals. A combined noise estimate based on the single and dual microphone noise estimates is then determined. Using the combined noise estimate, a gain mask may be generated and applied to the primary acoustic signal to generate a noise suppressed signal. Subsequently, the noise suppressed signal may be output.

System and method for providing single microphone noise suppression fallback

Systems and methods for utilizing inter-microphone level differences (ILD) to attenuate noise and enhance speech are provided. In exemplary embodiments, primary and secondary acoustic signals are received by omni-directional microphones, and converted into primary and secondary electric signals. A differential microphone array module processes the electric signals to determine a cardioid primary signal and a cardioid secondary signal. The cardioid signals are filtered through a frequency analysis module which takes the signals and mimics a cochlea implementation (i.e., cochlear domain). Energy levels of the signals are then computed, and the results are processed by an ILD module using a non-linear combination to obtain the ILD. In exemplary embodiments, the non-linear combination comprises dividing the energy level associated with the primary microphone by the energy level associated with the secondary microphone. The ILD is utilized by a noise reduction system to enhance the speech of the primary acoustic signal.

System and method for utilizing omni-directional microphones for speech enhancement

Author(s): Algazi, Ralph; Avendano, Carlos; Duda, Richard O | Abstract: An optimal spherical-head model is derived from the high-frequency interaural time difference (ITD) of a population of 25 subjects. Analysis demonstrates that the optimal sphere results in a very small lateral angle errors, except near5 the interaural poles. The spherical-head model is then estimated using the anthropometry of the subjects, based on simple and robust empirical predictive equations. Such a customization greatly decreases objective angular errors that occur when a generic model is used.

https://escholarship.org/content/qt3h41r825/qt3h41r825.pdf?t=ptt2wl

Carlos Avendano

Papers

The CIPIC HRTF database

Elevation localization and head-related transfer function analysis at low frequencies.

System and method for providing single microphone noise suppression fallback

System and method for utilizing omni-directional microphones for speech enhancement

Estimation of a Spherical-Head Model from Anthropometry