Acoustic Impulse Response Research Articles

Archaeoacoustic Examination of Lazarica Church

The acoustic analysis provides additional information on building tradition and related indoor practice that includes sound, thus deepening our understanding of architectural heritage. In this paper, the sound field of the Orthodox medieval church Lazarica (Kruševac city, Serbia) is examined. Lazarica is a representative of Morava architectural style, developed in the final period of the Serbian medieval state, when also the chanting art thrived, proving the importance of the aural environment in Serbian churches. The church plan is a combination of a traditional inscribed cross and a triconch. After the in situ measurement of acoustic impulse response using EASERA software, we built a computer model in the acoustic simulation software EASE and calibrated it accordingly. Following the parameters (reverberation time (T30), early decay time (EDT) and speech transmission index (STI)), we examined the acoustic effect of the space occupancy, central dome and the iconostasis. In all the cases, no significant deviation between T30 and EDT parameter was observed, which indicates uniform sound energy decay. Closing the dome with a flat ceiling did not show any significant impact on T30, but it lowered speech intelligibility. The height of iconostasis showed no significant influence on the acoustics of Lazarica church.

Discrete material optimization for wave-based acoustic design

The problem of automatic design of acoustic spaces is prevalent in architecture and room acoustics. We present a novel algorithm to automatically compute the optimal materials of large architectural spaces. Our method uses discrete optimization techniques to determine the best material configuration for desired acoustic properties of a structure, while taking into account properties of real-world materials. An efficient acoustic wave solver is used to accurately compute the acoustic impulse responses that drive the optimization process. Our method is tested on various computer representations of real-world scenes where we show how new material characteristics can be computed to improve the scene's strength, clarity, and reverberation time.

Fast Algorithms for Acoustic Impulse Response Shaping

Impulse response shaping (IRS) is a prefiltering technique for modifying the characteristics of a linear channel to achieve desirable characteristics. Acoustic impulse response shaping is used to apply partial equalization to reduce the effects of reverberation on audio signals propagating inside a room and is thus used for listening room compensation. To be capable of responding rapidly to time varying room acoustics, a shaping algorithm must be computationally fast; this is difficult to achieve since the filters to be designed may have thousands of coefficients. This paper derives variants of the Dual Augmented Lagrangian Method, presents their application to the design of the shaping filters, and validates their efficacy through experimental studies. In addition, the formulations have been modified to design shaping filters that are robust to change in microphone positions.

Influence of postharvest water loss on apple quality: The use of a sensory panel to verify destructive and non-destructive instrumental measurements of texture

‘Cripps Pink’/Pink Lady® and ‘Sciros’/Pacific Rose™ apples were stored at 0.5 °C in containers through which dry or humidified air was passed for up to 19 weeks. Sensory analysis by trained panellists was undertaken on apples that were removed from storage at a single point in time (15 or 19 weeks, respectively) so that cultivars and treatments could be directly compared. Weight loss during storage was 1.8 to 2.4 % for humidified air treatments and 5.0 to 5.7% for dry air treatments. The trained panel perceived differences between cultivars, but within cultivars there was no significant sensory difference between apples from higher and lower water loss treatments, except for juiciness which was lower for dehydrated apples. Measurements of puncture firmness were in accordance with the sensory results, suggesting that there was no effect of water loss on mechanical properties. However, non-destructive measurements (based on acoustic impulse and tissue impact responses) declined with water loss. These sensory-instrumental data along with monitoring during storage suggested that non-destructive measurements, particularly those based on acoustic impulse responses, are detecting changes in water status rather than mechanical properties of the fruit.

Linear System Identification Based on a Kronecker Product Decomposition

Linear system identification is a key problem in many important applications, among which echo cancelation is a very challenging one. Due to the long length impulse responses (i.e., echo paths) to be identified, there is always room (and needs) to improve the performance of the echo cancelers, especially in terms of complexity, convergence rate, robustness, and accuracy. In this paper, we propose a new way to address the system identification problem (from the echo cancelation perspective), by exploiting an optimal approximation of the impulse response based on the nearest Kronecker product decomposition. Also, we make a first step toward this direction, by developing an iterative Wiener filter based on this approach. As compared to the conventional Wiener filter, the proposed solution is much more attractive since its gain is twofold. First, the matrices to be inverted (or, preferably, linear systems to be solved) are smaller as compared to the conventional approach. Second, as a consequence, the iterative Wiener filter leads to a good estimate of the impulse response, even when a small amount of data is available for the estimation of the statistics. Simulation results support the theoretical findings and indicate the good results of the proposed approach, for the identification of different network and acoustic impulse responses.

Characterizing acoustic environments using spherical loudspeaker and microphone arrays

Room acoustic parameters, including the reverberation time (T60) and the direct-to-reverberant ratio (DRR), quantitatively describe the behaviour of an acoustic environment. These parameters are typically derived from an acoustic impulse response (AIR) obtained using an omnidirectional sound source and an omnidirectional receiver. However, the acoustic interaction between a source with a complex radiation pattern, for example, a human talker, and a receiver with a complex directivity pattern, for example, a human listener, cannot be accurately described by an omnidirectional AIR or the acoustic parameters derived from it. Here, we propose characterizing acoustic environments while taking the source and receiver directionality into account. A spherical loudspeaker array allows modelling a source with an arbitrary radiation pattern. Analogously, a spherical microphone can be used to describe a receiver with an arbitrary directivity pattern. By combining spherical loudspeaker and microphone arrays, a multiple-input multiple-output (MIMO) AIR of an acoustic environment can be measured to simulate sources and receivers with arbitrary directivities and estimate acoustic parameters accordingly.

Proof of concept: Machine learning based filling level estimation for bulk solid silos

Rigid silos are often under pressure and filled with hazardous materials. Therefore, in very few cases their level can be checked visually. For example, the level of mobile bulk solids silos, which are often used on construction sites, is checked by a worker throwing a stone against the silo and uses the sound to estimate the invisible silo cavity. This method is subjective and shows great errors based on experience. The proposed method should implement this principle robustly by machine application. A sensor unit is to give a mechanical impulse to the outer silo wall with a percussion mechanism and record the resulting sound with a microphone. Just as a human being is able to make an experience-based statement about the filling level of the silo over the growing amount of sounds heard, different machine learning approaches are to be considered in order to imitate this procedure. Machine learning approaches usually aim to find patterns between different sizes. A strongly approximated mathematical model will be used to prove that there is a noisy but demonstrable direct correlation between the level and the acoustic impulse response. This should justify future efforts to find suitable machine learning methods for this application.

Iterative echo labeling technique for room geometry estimation

Room geometry estimation using acoustic room impulse responses (RIRs) has been tackled in various ways. Most of them detect wall locations using the time of arrival (TOA) of echoes between multiple microphones and loudspeakers. To combine TOA information from multiple room impulse responses, echoes from the same walls should be collected and processed together. When multiple microphones are widely distributed in a room, however, it is hard to identify walls responsible for the generation of each echo. In this work, an iterative update algorithm is proposed for resolving the echo labeling problem with low computational complexity. Unlike conventional algorithms utilizing the rank property of Euclidean distance matrix, the proposed algorithm utilizes the property of convex hull formed by ellipses representing individual TOAs. The location of the reflectors are sequentially identified from common tangent lines constituting the convex hull, and the search result is iteratively updated by inspecting later echoes in time. In addition, a termination criterion is also developed in order to enable the geometry estimation without the prior knowledge of the number of the reflectors.

Modeling of the power delay profile for underwater acoustic channels—Effects of out-of-plane scattering and reverberation

Long reverberation tails seem to be a common feature of shallow water acoustic channel impulse responses. We show examples from two significantly different environments along the Norwegian coast, and demonstrate that 2-D propagation modeling fails to reproduce the long tails observed. The continuous nature of the impulse responses and the low correlation and slow decay rates of the tails point towards reverberation as an important mechanism behind the observations. The effects of reverberation and out-of-plane scattering are then included in the modeling by employing the 3-D ray-based model Rev3D. Scattering from the sea surface and sea bottom are described by scattering-strength functions including azimuthal variation. The modeling results indicate that the long reverberation tails can be explained by 3-D out-of-plane scattering. Although the measured levels are not perfectly matched, the importance of including reverberation in the modeling is clearly demonstrated.

Exact solution for the acoustical impulse response of a line source above an absorbing plane.

The solution of the two-dimensional (2D) wave equation excited by a Dirac delta function above an infinite plane is the 2D impulse response of the half-space. Equivalently, it can be considered as the sound field generated by an infinite line source in three-dimensional (3D) space. For a point source in 3D, it is possible to express the resulting impulse function in closed form if the infinite plane is characterized by a masslike or pure absorbing behavior. Based on the 3D impulse response above an absorbing plane, a closed form solution for a line source over absorbing ground is found. This solution consists of the direct pressure contribution plus a reflected signal. The analytical expression for the reflection factor R is the main result of the work. R is a simple algebraic function of impedance, time, and spatial coordinates. The acoustical field generated by a line source with a general time dependence can be constructed by convolution. For a Gaussian signal, the resulting pressure is computed numerically. For a rectangular pulse, an analytical solution in 3D and 2D is derived. The deviations from the numerically obtained results are very small.

Multi-Speaker DOA Estimation Using Deep Convolutional Networks Trained With Noise Signals

Supervised learning-based methods for source localization, being data driven, can be adapted to different acoustic conditions via training and have been shown to be robust to adverse acoustic environments. In this paper, a convolutional neural network (CNN) based supervised learning method for estimating the direction of arrival (DOA) of multiple speakers is proposed. Multi-speaker DOA estimation is formulated as a multi-class multi-label classification problem, where the assignment of each DOA label to the input feature is treated as a separate binary classification problem. The phase component of the short-time Fourier transform (STFT) coefficients of the received microphone signals are directly fed into the CNN, and the features for DOA estimation are learnt during training. Utilizing the assumption of disjoint speaker activity in the STFT domain, a novel method is proposed to train the CNN with synthesized noise signals. Through experimental evaluation with both simulated and measured acoustic impulse responses, the ability of the proposed DOA estimation approach to adapt to unseen acoustic conditions and its robustness to unseen noise type is demonstrated. Through additional empirical investigation, it is also shown that with an array of M microphone our proposed framework yields the best localization performance with M-1 convolution layers. The ability of the proposed method to accurately localize speakers in a dynamic acoustic scenario with varying number of sources is also shown.

Linear Prediction-Based Online Dereverberation and Noise Reduction Using Alternating Kalman Filters

Multichannel linear prediction-based dereverberation in the short-time Fourier transform (STFT) domain has been shown to be highly effective. Using this framework, the desired dereverberated multichannel signal is obtained by filtering the noise-free reverberant signals using the estimated multichannel autoregressive (MAR) coefficients. To use such methods in the presence of noise, especially in the case of online processing, remains a challenging problem. Existing sequential enhancement structures, which first remove the noise and then estimate the MAR coefficients, suffer from a causality problem as both the optimal noise reduction and dereverberation stages depend on the current output of each other. To address this problem, an algorithm that consists of two alternating Kalman filters to estimate the noise-free reverberant signals and the (MAR) coefficients is proposed. The causality of the estimation procedure is important when dealing with time-variant acoustic scenarios, where the MAR coefficients are time-varying. The proposed method is evaluated using simulated and measured acoustic impulse responses and is compared to a method based on the same signal model. In addition, a method to control the reverberation reduction and noise reduction independently is derived.

Efficient projection onto a low-dimensional ball

ABSTRACTProjection onto convex sets is a technique well known in optimization for its versatility and wide range of applications. This article presents an efficient projection onto a low-dimensional ball. The technique is based on analytically solving a quartic equation, and is exceptionally computationally efficient. An example application of this projection technique relates to acoustic impulse response shaping. Acoustic impulse response shaping is a pre-filtering technique to reduce reverberation of sound reproduction inside a room. If the focus is solely on the time domain, peaks and troughs in the frequency domain can occur. The projection approach presented in this article has been combined with the shaping algorithm to flatten the frequency response, thus providing control of both time domain and frequency domain characteristics.

Comparative study of two geometrical acoustic simulation models

Simulations based on the concepts of geometrical acoustics are today well-established tools for acousticians, being widely used for evaluation of sound quality in rooms and urban spaces. However, although a lot of different models are available and have been evaluated in the past, it is still very important to guarantee the validity and quality of simulated data and reproduced sound. This work presents a comparison between the signal processing strategies in two acoustic simulators based on geometrical models. Obvious expectation was that both simulators would produce the same results when fed by exactly the same input data. However, issues related to model assumptions, propagation methods characteristics and signal processing techniques adopted by each simulator introduce differences which alter the final results, i.e., the simulated acoustic impulse responses. This papers aims to present such deviation and helps to understand the influence of each component over the results. Firstly, both simulators are described in detail, presenting their acoustic models and the signal processing approaches. In addition, an extensive analysis of early reflections is performed, considering pressure levels, reflection order, their arrival time and directional characteristics. Next, simulated energy decay curves, monaural room acoustic parameters and spectra are objectively compared to measured data of a reverberant chamber, in two different conditions. The differences are then pointed out and minimized by unifying the signal processing of both simulators. The results of this comparison reveal that signal processing and inherit method characteristics still have strong influence over the simulated impulse responses, mainly for the late part. Some consequences are energy misbalance between early and late parts of impulse response, leading to differences over the room acoustic parameters, mainly clarity and definition.

Research on pipeline blocking state recognition algorithm based on mixed domain feature and KPCA-ELM

Aiming at the problem of recognition on pipeline blockage, a method based on mixed domain feature and KPCA-ELM is proposed. Firstly, the original acoustic impulse response signals are analysed by statistical analysis and local mean decomposition (LMD), in order to construct the mixed domain features, which are made up of time, frequency and time-frequency domain features. Then the kernel principal component analysis (KPCA) is adopted to reduce the high-dimensional features of mixed domain and extract the main features which reflect the operation state of main components. Finally, the main features are input to extreme learning machine (ELM) for state recognition. After the feature extraction by KPCA, the redundancy of input features is eliminated. The simulation results show that KPCA is more sensitive to the nonlinear characteristics of the pipeline blockage signal when compared with PCA. Meanwhile, ELM is superior to BP in terms of classification accuracy and time consuming.

Machine learning the acoustic impulse response of a rough surface

Due to time constraints imposed by operational considerations, reverberation models used by the Navy split the computational work by modeling the propagation and the scattering separately. The scattering module is typically a “look-up” table linking the incident sound field at any point on the boundary to a pre-computed statistic of the scattering. There are at least two significant drawbacks to this standard modeling approach. First, each acoustic interaction with the boundary generates a unique realization of a stochastic process, and consequently, no single statistic of the scattering always can be justified and the correct modeling of the statistics of reverberation time series must acknowledge this event-to-event variation. Second, pre-computing the scattering based on an assumed statistical description of the unresolved boundary irregularities without consideration of the resolved boundary irregularities prohibits the coherent inclusion in that scattering calculation of the effects that the resolved boundary irregularities have on the scattering. Machine-learned models for the boundary scattering hold the promise of addressing both of these issues. The purpose of this talk is to explore these issues and the prospect of developing a machine-learned model for the acoustic impulse response of a multi-scale rough ocean boundary. [This work was supported by the U.S. Office of Naval Research.]

Watermelon ripeness detection by wavelet multiresolution decomposition of acoustic impulse response signals

Ripeness is an important index for evaluating the quality of watermelons. This paper presents a simple, effective, and inexpensive approach to detecting watermelon ripeness based on wavelet multi-resolution decomposition (WMRD) and a statistical hypothesis test. The key idea in the proposed method is to model each WMRD coefficient as a random variable. For a specific coefficient, two samples are obtained by the multi-scale decomposition of watermelon acoustic signals, where one is for ripe watermelons and the other is for unripe ones. It is found that that many coefficients can be regarded as normal random variables, both for ripe and unripe watermelons. A discrimination index is proposed to estimate the separation degree of two samples, in order to determine the best WMRD coefficient. The probability density functions for ripe and unripe watermelons are obtained using the samples with the best coefficient. As a result, an inequality based on the hypothesis test at a 5% significant level is established for detecting watermelon ripeness. Using the proposed method, the training and test accuracies can reach 91.67% and 91.76%, respectively.

A Statistically Robust Approach to Acoustic Impulse Response Shaping

Acoustic impulse response shaping is a prefiltering technique to reduce the perceptible effects of room reverberation. The room impulse responses to be shaped must first be measured, and these measurements can contain errors. Furthermore, room responses vary with changes in temperature and humidity and also with changes in measurement position. This letter presents a method for enhancing the robustness of the shaping procedure so that the effects of these errors and changes are minimized. The method uses a stochastic model of the channel variations to explicitly limit the probability of large deviations from the desired performance. The method is evaluated on realistic channel perturbations and the resulting shaped responses are shown to comply with the robustness specification.

Spectrum and Energy Efficient OFDM Multicarrier Modulation for an Underwater Acoustic Channel

Time domain synchronization orthogonal frequency division multiplexing technique (TDS-OFDM) can offer better efficiency in spectrum and energy use than standard cyclic prefix and zero padding schemes by replacing the guard interval with a pseudorandom noise sequence. In this paper, a type of frequency domain pseudorandom noise training sequence is proposed for underwater acoustic channel as a TDS-OFDM training sequence and used to update the Doppler scaling factor estimation obtained by using a preamble signal after the initial system is synchronized. A significant improvement can also be achieved in bit error rate performance by the use of the compressed sensing theory to estimate underwater acoustic channel impulse response based on the proposed TDS-OFDM training sequence. Furthermore, the guard interval amplitude in TDS-OFDM can be reduced to improve the resultant energy efficiency and signal-to-interfered-signal ratio. The proposed techniques are carefully developed based on intuitive analysis and combined in a receiver design. Such receiver is evaluated in a measured underwater acoustic channel and a simulated channel. Both technical analysis and experiment results show significant and consistent performance improvements in spectral and energy use.

Non-destructive Quality Monitoring of Fresh Fruits and Vegetables

Quality determines the shelf life as well as selling price of fresh fruit or vegetable and therefore, quality monitoring and testing of fresh commodities have paramount importance in their postharvest handling and supply chain management. Most of the methods used to assess fruits and vegetables quality are destructive in nature. Now-a-days, various mechanical, optical, electromagnetic, and dynamic non-destructive methods are gaining importance due to ease in operations, faster turn over and reliability. Some of the non-destructive techniques (NDT) are currently being used in laboratories, research institutions and food packaging and processing industries, whereas, some methods are still at developmental stage. Various NDT with respect to their principle and applications such as impact test, electronic nose, time-resolved reflectance spectrometry (TSR), near infrared spectroscopy (NIR), nuclear magnetic resonance (NMR), X-Ray, ultra sonic, acoustic impulse response method, electrical conductivity methods etc., are discussed in this review.