Martin Hansen

Continuous assessment of time-varying speech quality.

This paper addresses the question of whether subjects are able to assess the perceived time-varying quality of speech material continuously. A method is introduced which is characterized by a subjective continuous rating of the perceived speech quality by moving a slider along a graphical scale. The usability of this method is illustrated with an experiment in which different sequences of sentences were degraded in quality with a Modulated Noise Reference Unit. The modulation depth was varied with time and the subjects task was to assess the perceived quality. The results indicate that subjects can monitor speech quality variations very accurately with a delay of approximately 1 s. An objective speech quality measure based on an auditory processing model was applied to predict the subjective speech quality results. The speech quality measure qC was modified to allow for time-dependent objective measurement of the speech quality. The averaged subjective response data could be modeled by the scale transformed and low-pass filtered measure qC(t) with a high degree of accuracy.

Using a quantitative psychoacoustical signal representation for objective speech quality measurement

This paper describes the application of a quantitative psychoacoustical signal preprocessing model for objective speech quality measurement. The preprocessing is applied to transform the original and the distorted speech signal to an internal representation which is thought of as the information that is accessible to higher neural stages of perception. From a comparison of these internal representations a quality measure can be derived that shows a high correlation to the subjective MOS data of various test databases. The inherent parameters of the preprocessing model were derived directly from psychoacoustical data independent of the present study. The detection thresholds of codec-like distortions obtained in a psychoacoustical experiment could also be predicted by the model. This indicates that the internal representation contains the relevant information for detecting perceivable differences. It provides evidence for a direct relation between speech quality and detectability of a distortion.

Objectively measured and subjectively perceived distortion in nonlinear systems

A method for measuring nonlinear distortion, which is applicable to almost any transmission system and which can use any input signal as a test signal, is proposed. The method exploits the fact that a pair of signals, generated to form the real and imaginary parts of the analytical signal corresponding to the input signal to a system, lose their property of being a Hilbert pair after being passed through a nonlinear system. The method was tested by measuring 12 different hearing aid compression systems. These objective measurements were compared with the subjectively perceived amount of distortion, assessed by a group of 10 otologically normal subjects. A reasonably monotonic relation between the subjective and objective measures of distortion was observed. The objective Hilbert pair based measure can be related to both traditional total harmonic distortion and traditional signal to noise ratio.

Smoothing individual head-related transfer functions in the frequency and spatial domains

When re-synthesizing individual head related transfer functions (HRTFs) with a microphone array, smoothing HRTFs spectrally and/or spatially prior to the computation of appropriate microphone filters may improve the synthesis accuracy. In this study, the limits of the associated HRTF modifications, until which no perceptual degradations occur, are explored. First, complex spectral smoothing of HRTFs into constant relative bandwidths was considered. As a prerequisite to complex smoothing, the HRTF phase spectra were substituted by linear phases, either for the whole frequency range or above a certain cut-off frequency only. The results indicate that a broadband phase linearization of HRTFs can be perceived for certain directions/subjects and that the thresholds can be predicted by a simple model. HRTF phase spectra can be linearized above 1 kHz without being detectable. After substituting the original phase by a linear phase above 5 kHz, HRTFs may be smoothed complexly into constant relative bandwidths of 1/5 octave, without introducing noticeable artifacts. Second, spatially smoother HRTF directivity patterns were obtained by levelling out spatial notches. It turned out that spatial notches do not have to be retained if they are less than 29 dB below the maximum level in the directivity pattern.

Regularization approaches for synthesizing HRTF directivity patterns

As an alternative to traditional artificial heads, it is possible to synthesize individual head-related transfer functions (HRTFs) using a so-called virtual artificial head (VAH), consisting of a microphone array with an appropriate topology and filter coefficients optimized using a narrowband least squares cost function. The resulting spatial directivity pattern of such a VAH is known to be sensitive to small deviations of the assumed microphone characteristics, e.g., gain, phase and/or the positions of the microphones. In many beamformer design procedures, this sensitivity is reduced by imposing a white noise gain (WNG) constraint on the filter coefficients for a single desired look direction. In this paper, this constraint is shown to be inappropriate for regularizing the HRTF synthesis with multiple desired directions and three alternative different regularization approaches are proposed and evaluated. In the first approach, the measured deviations of the microphone characteristics are taken into account in the filter design. In the second approach, the filter coefficients are regularized using the mean WNG for all directions. The third approach additionally takes into account several frequency bins into both the optimization and the regularization. The different proposed regularization approaches are compared using analytic and measured transfer functions, including random deviations. Experimental results show that the approach using multiple frequency bands mimicking the spectral resolution of the human auditory system yields the best robustness among the considered regularization approaches.

Pitch estimation of stereophonic mixtures of delay and amplitude panned signals

In this paper, a novel method for pitch estimation of stereophonic mixtures is presented, and it is investigated how the performance is affected by the pan parameters of the individual signals of the mixture. The method is based on a signal model that takes into account a stereophonic mixture created by mixing multiple individual channels with different pan parameters, and is hence suited for use in automatic music transcription, source separation and classification systems. Panning is done using both amplitude differences and delays. The performance of the estimator is compared to one single-channel, two multi-channel and one multi-pitch estimator using synthetic and real signals. Experiments show that the proposed method is able to correctly estimate the pitches of a mixture of three real signals when they are separated by more than 25 degrees.

Modeling speech intelligibility and quality on the basis of the ‘‘effective’’ signal processing in the auditory system

A model of the ‘‘effective’’ signal processing performed by the auditory system is applied to the quantitative assessment of speech intelligibility (cf., Kollmeier et al., Proc. ICA 95, III, 81–84). It consists of a gamma one filterbank followed by envelope extraction, nonlinear adaptation circuits, and subsequent modulation low‐pass filtering in each filter channel before being fed into a cross‐correlation‐type detector or pattern recognizer. The model is capable of predicting the speech intelligibility of both normal and hearing‐impaired listeners under a wide range of signal‐to‐noise ratios, as well as the perceived quality of speech signals whose spectro‐temporal properties are systematically degraded. The degradations are generated with a modulated noise reference unit either in a band‐specific, time‐independent way or in a broadband, temporally varying way. These intelligibility and quality assessment data can be used to deduce the components of the model most germane to the processing of speech, as...

Multi-pitch estimation of audio recordings using a codebook-based approach

In this paper, a method for multi-pitch estimation of single-channel mixtures of harmonic signals is presented. Using the method, it is possible to resolve amplitudes of overlapping harmonics, which is otherwise an ill-posed problem. The method is based on the extended invariance principle (EXIP), and a codebook consisting of realistic amplitude vectors. A nonlinear least squares (NLS) cost function is formed based on the observed signal and a parametric model of the signal, for a set of fundamental frequency candidates. For each of these, amplitude estimates are computed. The magnitudes of these estimates are quantized according to a codebook, and an updated cost function is used to estimate the fundamental frequencies of the sources. The performance of the proposed estimator is evaluated using synthetic and real mixtures, and the results show that the proposed method is able to estimate multiple pitches in a mixture of sources with overlapping harmonics.

Silicon cochlea: A digital VLSI implementation of a quantitative model of the auditory system

In this paper a digital VLSI implementation of a quantitative model of the auditory system is presented, which includes several processing stages physiologically and psychoacoustically motivated by the function of the human ear. The model was successfully applied to a wide range of applications from psychoacoustical experiments to speech recognition. The hardware design is derived from a C/C++ notation of the algorithms using floating point arithmetics. One application of the model is used to determine the necessary internal precision of the computations for a transfer into a version suitable for a hardware implementation using fixpoint arithmetics. This application provides a very valid test bench since the significant speech processing features of the model are proofed by computing an objective speech quality measure. The processing scheme is divided into designs for two FPGAs and will be converted into ASICs in a later version. The implementation of the model as dedicated hardware provides efficient co...

Continuous assessment and modeling of speech transmission quality

This paper addresses the question of whether subjects are able to assess the perceived time‐varying quality of speech material continuously. A method inspired by Hamberg and de Ridder [J. Opt. Soc. Am. A 12, 2573–2577 (1995)] is presented by which subjects continuously rate the perceived speech quality by moving a slider along a graphical scale. In this way, temporal variations in quality can be monitored quantitatively and differences between, for example, alternative speech transmission systems, can be analyzed in an informative way. The usability of this method is illustrated with an experiment in which different sequences of sentences with a total duration of 40 s were degraded in quality with a modulated noise reference unit. The modulation depth was varied with time and the subjects task was to assess the perceived quality. The results show that subjects can monitor speech quality variations closely with a delay of approximately 1 s. The subjective results are discussed in view of a psychoacoustical...