Properties Of Human Hearing Research Articles

A convolutional neural-network model of human cochlear mechanics and filter tuning for real-time applications.

Auditory models are commonly used as feature extractors for automatic speech-recognition systems or as front-ends for robotics, machine-hearing and hearing-aid applications. Although auditory models can capture the biophysical and nonlinear properties of human hearing in great detail, these biophysical models are computationally expensive and cannot be used in real-time applications. We present a hybrid approach where convolutional neural networks are combined with computational neuroscience to yield a real-time end-to-end model for human cochlear mechanics, including level-dependent filter tuning (CoNNear). The CoNNear model was trained on acoustic speech material and its performance and applicability were evaluated using (unseen) sound stimuli commonly employed in cochlear mechanics research. The CoNNear model accurately simulates human cochlear frequency selectivity and its dependence on sound intensity, an essential quality for robust speech intelligibility at negative speech-to-background-noise ratios. The CoNNear architecture is based on parallel and differentiable computations and has the power to achieve real-time human performance. These unique CoNNear features will enable the next generation of human-like machine-hearing applications.

Longitudinal spread of mechanical excitation through tectorial membrane traveling waves

The mammalian inner ear separates sounds by their frequency content, and this separation underlies important properties of human hearing, including our ability to understand speech in noisy environments. Studies of genetic disorders of hearing have demonstrated a link between frequency selectivity and wave properties of the tectorial membrane (TM). To understand these wave properties better, we developed chemical manipulations that systematically and reversibly alter TM stiffness and viscosity. Using microfabricated shear probes, we show that (i) reducing pH reduces TM stiffness with little change in TM viscosity and (ii) adding PEG increases TM viscosity with little change in TM stiffness. By applying these manipulations in measurements of TM waves, we show that TM wave speed is determined primarily by stiffness at low frequencies and by viscosity at high frequencies. Both TM viscosity and stiffness affect the longitudinal spread of mechanical excitation through the TM over a broad range of frequencies. Increasing TM viscosity or decreasing stiffness reduces longitudinal spread of mechanical excitation, thereby coupling a smaller range of best frequencies and sharpening tuning. In contrast, increasing viscous loss or decreasing stiffness would tend to broaden tuning in resonance-based TM models. Thus, TM wave and resonance mechanisms are fundamentally different in the way they control frequency selectivity.

Automatic speech signal segmentation based on the innovation adaptive filter

Abstract Speech segmentation is an essential stage in designing automatic speech recognition systems and one can find several algorithms proposed in the literature. It is a difficult problem, as speech is immensely variable. The aim of the authors’ studies was to design an algorithm that could be employed at the stage of automatic speech recognition. This would make it possible to avoid some problems related to speech signal parametrization. Posing the problem in such a way requires the algorithm to be capable of working in real time. The only such algorithm was proposed by Tyagi et al., (2006), and it is a modified version of Brandt’s algorithm. The article presents a new algorithm for unsupervised automatic speech signal segmentation. It performs segmentation without access to information about the phonetic content of the utterances, relying exclusively on second-order statistics of a speech signal. The starting point for the proposed method is time-varying Schur coefficients of an innovation adaptive filter. The Schur algorithm is known to be fast, precise, stable and capable of rapidly tracking changes in second order signal statistics. A transfer from one phoneme to another in the speech signal always indicates a change in signal statistics caused by vocal track changes. In order to allow for the properties of human hearing, detection of inter-phoneme boundaries is performed based on statistics defined on the mel spectrum determined from the reflection coefficients. The paper presents the structure of the algorithm, defines its properties, lists parameter values, describes detection efficiency results, and compares them with those for another algorithm. The obtained segmentation results, are satisfactory.

The audibility of direct sound as a key to measuring the clarity of speech and music

Human ear/brain systems evolved to decode the direction, timbre, and distance of multiple sounds in a complex and noisy environment. In a reverberant space, this information is only available at the onset of a sound, before reflections overwhelm it. But since the time of Sabine acoustic science has concentrated on the decay of sound in a reverberant field, not on the audibility of the onset information. In addition, it is well known that the ability to separate multiple sound sources depends critically on pitch, but acoustic research studies only noise and impulses. This paper proposes that clarity requires the ability to separately analyze multiple sounds (the cocktail party effect) and that the cocktail party effect depends on phase relationships between harmonics of complex tones. These phase relationships are scrambled in predicable ways by reflections and reverberation. Well known properties of human hearing are used to develop both a physical model for the neurology of onset detection and an impulse response measure for localization and clarity in a reverberant field. A C language implementation of the physical model is capable of predicting and perhaps measuring the localizability of individual musicians in a binaural recording of live music.

Adapting noise measurement and modeling practices to meet legislative and policy mandates for National Park management.

Historical community noise studies focused on exposures corresponding to maximum tolerable annoyance and potential health risks. These criteria should rarely be pertinent for National Parks, but applicable knowledge and tools were developed by these studies. A-weighted measurements integrate sound energy across the audible spectrum to account for properties of human hearing. To improve the utility of A-weighted measurements, NPS is investigating band-limited dB(A) measures to exclude natural environmental sounds that have no bearing on the effects of anthropogenic noise. Leq metrics were shown to be reasonable for integrating time histories to predict human responses to noise exposure. NPS utilizes measurements of audibility to parse this measure into the extent of noise-free conditions and the Leq value when noise is audible. This approach distinguishes situations with chronic exposure to modest levels of noise from very infrequent exposures to loud events. Several noise models have been developed to support community noise management. These are used in park settings to model the kinds of individual noise events that can occur (a type of vehicle on a specific route). Interactive noise mapping tools have been developed to compute aggregate exposure from collections of many such events under different planning scenarios.

A Speech Enhancement Algorithm based on Human Psychoacoustic Property

In the speech system, for example hearing aid as well as speech communication, speech quality is degraded by environmental noise. In this study, to enhance the speech quality which is degraded by environmental speech, we proposed an algorithm to reduce the noise and reinforce the speech. The minima controlled recursive averaging (MCRA) algorithm is used to estimate the noise spectrum and spectral weighting factor is used to reduce the noise. And partial masking effect which is one of the human hearing properties is introduced to reinforce the speech. Then we compared the waveform, spectrogram, Perceptual Evaluation of Speech Quality (PESQ) and segmental Signal to Noise Ratio (segSNR) between original speech, noisy speech, noise reduced speech and enhanced speech by proposed method. As a result, enhanced speech by proposed method is reinforced in high frequency which is degraded by noise, and PESQ, segSNR is enhanced. It means that the speech quality is enhanced.

Energy first search scheme for spectral envelope vector quantisation of linear predictive coding residual

Linear predictive coding (LPC) residual is an important component for analysis-by-synthesis speech coders. Using spectral envelope vector quantisation (SEVQ) for LPC residual signals, MPEG-4 harmonic vector excitation coding (HVXC) speech coder is able to change the tone and speed for decoding bitstreams. A modified version is proposed to the simplified hearing-based SEVQ (SSEVQ) approach proposed by Wang and Yang. In addition, the energy first search scheme for SEVQ (EFSEVQ) is proposed in order to reduce the computational complexity of SEVQ for MPEG-4 HVXC speech coder. The proposed EFSEVQ scheme also agrees with human hearing properties. Simulation results reveal that the proposed EFSEVQ search scheme not only reduces the computational complexity but also preserves the quality of encoded speech.

Demonstration of properties of human hearing using a cochlear analogue

A device has been constructed that demonstrates the spectral decomposition of waves according to spatial location, analogous to the function of the basilar membrane of the mammalian cochlea. The design is based on a previous device constructed by Keolian [J. Acoust. Soc. Am. 101, 1199–1201 (1997)]. We have modified Keolian’s design to incorporate variable mass as well as stiffness and an electronically driven shaker, which allows the device to be driven with complex waveforms. In addition to showing spectral decomposition, the model can be used to demonstrate beating, masking, and other phenomena that occur on the basilar membrane. A demonstration of the device will be given.

Perceptual shape VQ of spectral envelope for efficient representation of LPC residual

The authors present an effective spectral envelope (SE) quantisation scheme for parametric speech coders, based on human hearing properties. The variable-dimension SE uniformly sampled vector in frequency is first converted into a fixed, but small, number of nonlinearly spaced frequency bands on the Bark scale. The minimum Bark spectral distortion (BSD) criterion is applied to enable the hearing-based SE vector quantisation (HSEVQ) scheme to quantise the SE vector, achieving a slightly better perceptual quality than the traditional method. A simplified HSEVQ (SSEVQ) scheme is developed by removing some of insensitive functions from the HSEVQ to reduce the complexity of the computation. Simulations reveal that the SSEVQ method reduces the amount of computation of the traditional SE vector quantisation scheme by a factor of nine, while retaining the quality of the reconstructed speech signal.

Incorporating the human hearing properties in the signal subspace approach for speech enhancement

The major drawback of most noise reduction methods in speech applications is the annoying residual noise known as musical noise. A potential solution to this artifact is the incorporation of a human hearing model in the suppression filter design. However, since the available models are usually developed in the frequency domain, it is not clear how they can be applied in the signal subspace approach for speech enhancement. In this paper, we present a Frequency to Eigendomain Transformation (FET) which permits to calculate a perceptually based eigenfilter. This filter yields an improved result where better shaping of the residual noise, from a perceptual perspective, is achieved. The proposed method can also be used with the general case of colored noise. Spectrogram illustrations and listening test results are given to show the superiority of the proposed method over the conventional signal subspace approach.

Parameter Estimation of a Plucked String Synthesis Model Using a Genetic Algorithm with Perceptual Fitness Calculation

We describe a technique for estimating control parameters for a plucked string synthesis model using a genetic algorithm. The model has been intensively used for sound synthesis of various string instruments but the fine tuning of the parameters has been carried out with a semiautomatic method that requires some hand adjustment with human listening. An automated method for extracting the parameters from recorded tones is described in this paper. The calculation of the fitness function utilizes knowledge of the properties of human hearing.

A new objective evaluation method of wind noise in a car based on human hearing properties.

This paper describes a new method for objective evaluation of wind noise in the passenger compartment of a car. The perception of wind noise is affected by the masking effect by other noise and the particular source direction of the wind noise. Therefore, the human hearing properties of masking and sound localization should be taken into account when evaluating wind noise. The proposed evaluation method is based on these human hearing properties. The method can estimate the loudness and direction of noise in each frequency band by performing a masked frequency spectrum analysis and a cross correlation analysis based on a binaural signal processing model. Therefore, those wind noise components that are annoying to the passengers or those wind noise components whose source location can be determined by the human listener can be identified objectively. Furthermore, the total loudness of wind noise as determined by a jury of human listeners can be estimated quite precisely by adding the loudness of the frequency bands for noise emanating from the direction of the side window.

Perceptual coding of wideband audio signals: Basic principles and current applications

The need for a significant reduction in bit rate for wideband digital audio signal transmission and storage has led in the last few years to the development of psychoacoustics-based data compression techniques. In this approach, the audio frequency range is subdivided into sub-bands which are approximations of human auditory critical bands. The frequency representation of audio signals is accomplished using a filter-bank implemented as a time-to-frequency transform or as a sub-band filter. Frequency-domain masking properties of the human auditory system are exploited in order to maximize perceived fidelity of a steady-state signal at a given bit rate. In a similar way, temporal masking effects are very important in the bit-rate reduction process of transient signals. In this talk, the basic principles of perceptual audio coding will be reviewed. An adaptive transform coding technique will be presented in which both frequency-domain and temporal masking properties of human hearing are exploited during bit-rate reduction. Dolby’s current and future applications of this coding technology will be discussed.

Ear‐adequate spectral analysis for noise evaluation

In numerous sound situations, sound quality cannot be determined by means of conventional measuring techniques. The frequent discrepancy between subjective judgment of sound events yielded by standard measurement procedures, can be explained by the difference between a sound situation as received by a single omnidirectional microphone, on the one hand, and as received by the human hearing, on the other hand. A new procedure using a dummy head and a special binaural analyzer has been developed for spectral analysis that realizes frequency and time resolution in the entire audio spectral range. This is comparable to the properties of human hearing, which is able to perceive quickly changing time structures and minor frequency differences at the same time. New approaches for ear‐adequate noise evaluation in order to judge noise quality can be achieved by means of psychoacoustic properties of human heating and binaural signal processing. The results of these investigations show that human hearing, which has two auditory paths with a complex directional characteristic and a binaural pattern recognition, gives a different judgment of noise quality as compared to conventional measurement methods, especially in complex sound situations with a spatial distribution of several sound sources and a sound‐pressure level below 85 dB.

Speculations about noise and the evolution of vertebrate hearing

Two of the most conspicuous and celebrated properties of human hearing — the abilities to discriminate pitch and to localize sound sources, may have been secondary derivations to which the ear was preadapted by having evolved in the inevitable presence of noise.