Hrvoje Domitrović

Estimating Perceptual Audio System Quality Using PEAQ Algorithm

The idea of objective, computer based methods for testing audio quality is introduced in this work. Including some basic of perceptual coding, international standards and description of new measurement algorithm. With described algorithm PEAQ we have measured a characteristics of four different codecs and have shown the results in this paper

Assessment of urban soundscapes with the focus on an architectural installation with musical features

Urban soundscapes at five locations in the city of Zadar were perceptually assessed by on-site surveys and objectively evaluated based on monaural and binaural recordings. All locations were chosen so that they would display auditory and visual diversity as much as possible. The unique sound installation known as the Sea Organ was included as an atypical music-like environment. Typical objective parameters were calculated from the recordings related to the amount of acoustic energy, spectral properties of sound, the amount of fluctuations, and tonal properties. The subjective assessment was done on-site using a common survey for evaluating the properties of sound and visual environment. The results revealed the importance of introducing the context into soundscape research because objective parameters did not show significant correlation with responses obtained from interviewees. Excessive values of certain objective parameters could indicate that a sound environment will be perceived as unpleasant or annoying, but its overall perception depends on how well it agrees with peoples expectations. This was clearly seen for the case of Sea Organ for which the highest values of objective parameters were obtained, but, at the same time, it was evaluated as the most positive sound environment in every aspect.

Student project of building an impedance tube

This paper describes a student project of building an impedance tube for measuring the absorption coefficient using the transfer‐function method, in accordance with the standard ISO 10534‐2. This method is well‐established and has many advantages compared to the older method using standing wave ratio (ISO 10354‐1) in terms of measurement speed and accuracy. For the tube, only inexpensive materials and transducers were used. The tube was designed for the frequency range between 90 and 2000 Hz. In order to achieve this range with one tube, three microphone positions have been used. The resulting absorption coefficient has been calculated using the one‐ and two‐microphone method. Different broadband excitation signals have been used in order to compare their robustness, such as MLS, frequency sweep and white noise. Various problems with the design and construction are addressed and the optimal configuration is discussed.

Reverberation time measuring methods

In this paper different well‐established methods of reverberation time measurement are compared. Furthermore, the results obtained using these methods are compared to the results provided by some additional methods which could serve as an in situ tool if, for any reason, the reverberation time measurements cannot be carried out using the standardized methods. The methods compared in this paper include the standardized methods (EN ISO 3382:2000), namely the impulse response measured with pink noise, exponential sweep, MLS, but also pistol shots of different calibers, balloon bursts, gated external pink noise, and the B&K filtered burst method. In order to make the comparison, the measurements were performed in four acoustically very different spaces ‐ a rather small and well‐damped listening room, a much bigger damped listening room, a rather reverberant atrium, and a large and very reverberant shoebox‐shaped room. The results were evaluated according to signal‐to‐noise ratio criterion as well. Special attention has been given to the influence of room modes on measurement results.

Sound Quality Evaluation of Hand-Held Power Tools

Sound quality evaluation in the context of product sound quality has been addressed and investigated on hand-held power tools. Out of those, power drills, hand-held circular saws and jigsaws have been chosen as the products of interest. A series of listening tests has been conducted in order to ascertain how people perceive not only the sound emitted by these devices in operation, but also certain properties of the devices themselves, such as their quality, safety of handling, robustness, proper functioning and others. The values of objective parameters describing the sounds of these devices were obtained. Well known parameters such as loudness, sharpness and roughness were used, and certain new parameters were defined. Linear regression models were made for subjective parameters describing the perception of the devices using relevant objective parameters as an attempt to model the subjective grades obtained through listening tests and to predict such grades for future products to be included in the investigation. Based on individual parameters, a single-number overall grade was defined.

A comparison of impulse‐like sources to be used in reverberation time measurements

As a part of an extensive ongoing research on reverberation time measurements, efforts have been made to find an impulse‐like source that will comply with certain demands concerning its spectral content and sound pressure levels it can provide. Although the relevant standards state that such kind of sources should be avoided due to the lack of measurement repeatability, their use has proved to be necessary due to inability of conventional omni‐directional loudspeakers to provide adequate sound pressure levels, especially at the low end of the frequency range of interest, thereby ensuring sufficient dynamic range. Therefore, the emphasis of this investigation has been made on peak sound pressure levels and the amount of low frequency content each source is able to provide. The investigated sources include 6 mm and 8 mm pistols, firecrackers with different amount of explosive compound and explosive mixtures of acetylene gas.

Measurements of nonlinear electroacoustical systems

When testing low bit-rate, i.e. audio systems comprising perceptual coding, one has to use test stimulus similar to real-world signals; on the other hand, it is necessary to use some kind of estimation of the temporal behaviour of human hearing process. This article shows the results we have obtained using our own developed test stimulus called CATS (composed audio test signal), and further analysis that gives transparency reduction as a measure of quality.

Improvement of acoustic resistance equations in perforated plate absorbers with thin porous layers

The absorption coefficient of perforated plate absorbers largely depends on the acoustic resistance of the backing layer, adjacent to the plate itself. The optimization process of Helmholtz resonators is focused on the maximization of their absorption coefficients, by tuning the acoustic resistance and reactance of the overall absorbing system. In this paper, the comparison of acoustic resistance of a perforated plate absorber for two characteristic cases is shown: one with a thin porous layer glued tightly to the backing of the perforated plate, and the other with a thin porous layer placed near the perforations, without tight contact. The measurements were done in the Kundts tube, by using four characteristic porosities of the plate (4.4%, 8.5%, 12.9% and 16.9%) and a 2 mm thick porous layer. The porous layer can be simply added to the acoustic resistance in the case of its tight placement to the perforated plate. On the other hand, absorber resistance is much less increased when a loosely placed porous layer is introduced to the perforations, but still shows a positive correlation with the resistance increase in the case of tight placement. Ratios of measured resistances are shown for the mentioned cases, and a range of useful values of those ratios is introduced as an improvement to the analytical expression used for calculating the acoustic impedance. A novel correction factor for calculating acoustic resistance of a perforated plate with porous layer in loose contact is proposed.

Frequency response measurement with composed audio test signal

The composed audio test signal (CATS) is formed by mixing three mutually independent composed saw signals (CSSs). The CSS sequences are a special type of pseudorandom signal. The CATS closely follows the characteristics of natural signals, and its behavior is controlled by the well‐defined and configurable parameters. The amplitude probability density function of the CATS follows Gaussian distribution. Furthermore, second and third CSS sequences are generated using the perturbated parameters of the first one. Pseudorandom noise can have a spectrum and amplitude distribution similar to that of white noise. The CATS methods employ efficient cross correlation between input and output to recover the periodic impulse response (PIR) of the system being measured. The PIR can be considered as the impulse response only if the CATSs period equals or exceeds the duration of the system impulse response. The frequency response can then be found by taking the discrete Fourier transform of the impulse response. The CATS...

Comparison of Statistical Model-Based Voice Activity Detectors for Mobile Robot Speech Applications

Abstract This paper deals with the problem of voice activity detection in adverse acoustic conditions, namely high and varying noise scenarios. For robotic applications, we need the voice activity detector to be computationally light, robust to varying levels of background noise, and have a low latency, especially if we are tracking moving speakers. We analyze three voice activity detectors—two model the discrete Fourier transform coefficients by Gaussian and generalized Gaussian distribution, while the third models the spectral envelope as having either Rayleigh or Rice distribution—and we present them in a unifying and consistent manner, with respect to a statistical hypotheses ratio measure and a joint noise spectrum estimation algorithm. Moreover, we compare the performance under various noise conditions; three types of noises, three different signal-to-noise ratios and six different speakers, by means of receiver operating characteristic curves and area under a curve score. The results showed that the Rayleigh-Rice model had on average better results and medium computational demand.