A Solution for Developing Corpora for Polish Speech Enhancement in Complex Acoustic Environments

Binaural hearing and speech communication often take place in complex acoustic environments (CAEs) with a variety of interfering sounds and reverberation. For hearing research and the evaluation of hearing systems, virtual CAEs gained in interest to approach ecological validity. Here, the effects of complexity and visual representation of the CAEs on sound source localization, distance perception, loudness, speech intelligibility, and listening effort were investigated in a virtual audio-visual environment. The simulated sound field was rendered using a 3D, 86-channel loudspeaker array in combination with or without a stereoscopic head-mounted display (HMD) for the visual representation. The visual scenes consisted of a ring of eight virtual loudspeakers in an anechoic and an echoic condition with a reverberation time of about 1.5 s. Complexity of the environment was varied through the number of interfering sound sources, the receiver-target distance and by assessing all psychoacoustic measures after a single presentation or in consecutive measurements with one presentation for each measure. No significant effect of wearing an HMD was found. Reverberation, the number of maskers, and distance had significant effects on all psychoacoustic measures. These factors also degraded auditory cues relevant in speech communication, such as interaural coherence. There were significant, however, generally small deviations between assessment of all psychoacoustic measures after a single or consecutive presentation. The suggested setup is suited to be directly transferred to a corresponding real room, enabling a $1: 1$ comparison between real and virtual environments.

Identification of bird species from audio recordings has been a major area of interest within the field of ecological surveillance and biodiversity conservation. Previous studies have successfully identified bird species from given recordings. However, most of these studies are only adaptive to low‐noise acoustic environments and the cases where each recording contains only one bird's sound simultaneously. In reality, bird audios recorded in the wild often contain overlapping signals, such as bird dawn chorus, which makes audio feature extraction and accurate classification extremely difficult. This study is the first to focus on applying a blind source separation method to identify all foreground bird species contained in overlapping vocalization recordings. The proposed IVA‐Xception model is based on independent vector analysis and convolutional neural network. Experiments on 2020 Bird Sound Recognition in Complex Acoustic Environments competition (BirdCLEF2020) dataset show that this model could achieve a higher macro F1‐score and average accuracy compared with state‐of‐the‐art methods.

The effect of healthy aging on change detection and sensitivity to predictable structure in crowded acoustic scenes

This paper describes acoustic models that can generate analytical metrics in support of naval operations, particularly in coastal oceans. Coastal environments are generally characterized by high spatial and temporal variabilities. When coupled with acoustic spectral dependencies of the surface and bottom boundaries, these natural variabilities make coastal regions very complex acoustic environments. Thus, accurate modeling of the acoustic environment is essential for prediction of sonar performance in coastal oceans. The current inventory of ocean-acoustic models comprises 126 propagation models, 19 noise models, 26 reverberation models and 34 sonar-performance models. Approximately 18 percent of this inventory is tailored for shallow-water applications. When coupled with coastal atmosphere-ocean models, these ocean-acoustic models can generate sophisticated prognostic and diagnostic metrics in support of naval operations in coastal oceans.

In experiments, control is everything. If parameters, procedures or the environment cannot be controlled thoroughly, any experiment lacks repeatability and hardly any findings can be generalized. In psychoacoustics and audiology research, important aspects of perception in real life are sacrificed in favor of control. The new “Real Life Lab” at Sonova allows audiological research to take a big step forward towards real life environments by bringing the freedom to move around in acoustic environments to controlled laboratory conditions. The lab provides a large stage where persons can move around freely, interact with acoustic or real objects, or with simulated or real people. The stage is surrounded by a framework of loudspeakers to present acoustic content from all directions and screens to present visual content in the horizontal plain. Any motion by persons on the stage can be tracked in real time with a motion capturing system. The motion capturing data can be passively tracked during any experiment or it can be actively used to alter or trigger audio and video reproduction. This contribution shows the technical capabilities of the lab together with some insights on the first experiment on the motion behavior in an orientation task in a complex acoustic environment.

When multiple acoustic sources are present in an environment monitored by multiple microphones, each microphone’s response signal is typically a weighted and delayed mixture of the active source signals. Subject to important constraints, these mixtures can be processed using Blind Source Separation (BSS) methods to construct output signals representative of each of the “hidden” acoustic sources in isolation. An important objective measure of the quality of a given “separation solution” is the amount of mutual information remaining in the BSS output channels: the less mutual information across channels, the better the separation. Often, the acoustic signals of interest in a complex acoustic environment are speech sources, and the measure of separation quality that really matters to listeners is speech intelligibility—a subjective (perceptual) measure that cannot be directly computed by a signal processing system. Thus, it would be useful to know how well objective measures of separation quality, such as mutual information, correlate with intelligibility. We present results of perceptual and objective tests exploring the relationship between objective separation metrics and word intelligibility as estimated by listeners’ responses in word identification tasks, for target-word-in-carrier-phrase utterances recorded in an acoustic environment with masking speech babble and broadband noise sources.

Target-speaker voice activity detection is currently a promising approach for speaker diarization in complex acoustic environments. This paper presents a novel Sequence-to-Sequence Target-Speaker Voice Activity Detection (Seq2Seq-TSVAD) method that can efficiently address the joint modeling of large-scale speakers and predict high-resolution voice activities. Experimental results show that larger speaker capacity and higher output resolution can significantly reduce the diarization error rate (DER), which achieves the new state-of-the-art performance of 4.55% on the VoxConverse test set and 10.77% on Track 1 of the DIHARD-III evaluation set under the widely-used evaluation metrics.

The ability to localize a sound source is very important in our daily life, specifically to analyze auditory scenes in complex acoustic environments. The concept of minimum audible angle (MAA), which is defined as the smallest detectable difference between the incident directions of two sound sources, has been widely used in the research fields of auditory perception to measure localization ability. Measuring MAAs usually involves a reference sound source and either a large number of loudspeakers or a movable sound source in order to reproduce sound sources at a large number of predefined incident directions. However, existing MAA test systems are often cumbersome because they require a large number of loudspeakers or a mechanical rail slide and thus are expensive and inconvenient to use. This study investigates a novel MAA test method using virtual sound source synthesis and avoiding the problems with traditional methods. We compare the perceptual localization acuity of sound sources in two experimental designs: using the virtual presentation and real sound sources. The virtual sound source is reproduced through a pair of loudspeakers weighted by vector-based amplitude panning (VBAP). Results show that the average measured MAA at 0° azimuth is 1.1° and the average measured MAA at 90° azimuth is 3.1° in a virtual acoustic system, meanwhile the average measured MAA at 0° azimuth is about 1.2° and the average measured MAA at 90° azimuth is 3.3° when using the real sound sources. The measurements of the two methods have no significant difference. We conclude that the proposed MAA test system is a suitable alternative to more complicated and expensive setups.

In a complex acoustic environment, several sound sources may simultaneously change their loudness, location, timbre, and pitch. Yet humans like many other animals are able to integrate effortlessly the multitude of cues arriving at their ears, and to derive coherent percepts and judgments about the different attributes of each source. This facility to analyze an auditory scene is conceptually based on a multi-stage process in which sound is first analyzed in terms of a relatively few perceptually significant attributes (the alphabet of auditory perception), followed by higher level integrative processes that organize and group the extracted attributes according to specific context-sensitive rules (the syntax of auditory perception) [1]. The sound received at the two ears is processed for attributes including source location, acoustic ambience, and source attributes such as tone and pitch, timbre and intensity.Decades of physiological and psychoacoustical studies [2,3] have revealed elegant strategies at various stages of the mammalian auditory system for the representation of the signal cues underlying auditory perception. This information has facilitated the development of biophysical models, mathematical abstractions, and computational algorithms of the early and central auditory stages with the aim of capturing the functionality, robustness, and enormous versatility of the auditory system [4]. Numerous groups have implemented such algorithms in software and hardware, and have evaluated them by comparing their performance to human performance and against a range of robustness and flexibility requirements. Furthermore, these auditory-inspired processing strategies have been utilized in a wide range of applications including acoustic diagnostic monitoring systems for machines and manufacturing processes, battlefield acoustic signal analysis, sound analysis and recognition systems, robust detection and recognition of multiple interacting faults, and detection and recognition of underwater transients and weak signals in low signal-to-noise ratio (SNR) in acoustically-cluttered environments [5,6].We shall briefly review the auditory encoding of various sound attributes to illustrate the above ideas. We shall specifically focus on the percept of sound timbre: what acoustic cues are most intimately correlated with it? How are they represented at various stages of the auditory pathway? And how the abstracted auditory signal processing algorithms and representations can be applied to measure speech intelligibility to describe musical timbre and to analyze complex auditory scenes?

The influence of the perceptual or fear learning on rats’ prepulse inhibition induced by changes in the correlation between two spatially separated noise sounds

Indoor speech propagation causes minute vibrations in surrounding objects, enabling remote speech recovery through passive eavesdropping. Unlike traditional methods that rely on acoustic waves, passive eavesdropping uses object vibrations, making it difficult to defend against, even in soundproof environments. However, weak vibration signals and noise interference make speech recovery challenging. Existing studies mainly focus on deep learning for signal reconstruction, requiring large datasets and high computational power, which complicates real-time, on-device deployment. To address this, we propose a lightweight passive speech recovery system based on millimeter-wave radar. Without prior knowledge of object locations or numbers, the system can adaptively fuse multi-source signals for real-time speech reconstruction. To counteract the noise characteristics of millimeter-wave radar and the weak amplitude of vibration signals, we designed a set of low-complexity noise suppression and signal enhancement algorithms, ensuring efficient operation on edge devices. Experimental results demonstrate that in single-target scenarios, the proposed system achieved a Mel Cepstral Distortion (MCD) of 3.923 and a Word Error Rate (WER) of 12.9%. In multi-target scenarios, the SNR improved by 3.65 dB, MCD decreased by an average of 1.52, and WER decreased by an average of 15.83%, making the method effective and practical in complex acoustic environments.

IntroductionAging is associated with decrements in speech-in-noise perception which make communication in real world environments difficult. However, the neural correlates of these difficulties are not well characterized and finding correlations between speech-in-noise performance and electrophysiological measures has been difficult due in part to the fact that speech-in-noise perception is a multi-faceted process. The current study used a wide range of speech-in-noise tasks in an attempt to more completely capture speech-in-noise performance and compared these with electrophysiological measures.MethodsP300 event related brain responses were elicited in young and older adult listeners to spoken isochronous syllable sequences presented in quiet and noisy (i.e., multi-talker babble) background conditions. To investigate the extent to which P300 responses are associated with speech-in-noise understanding, listeners also completed a separate battery of speech-in-noise recognition tasks.ResultsOverall P300 amplitudes measured in noisy conditions, but not in quiet conditions, were associated with performance on a variety of speech recognition tasks and were positively correlated with a composite measure of speech understanding in noise based on the full battery. In addition, older adults had P300 responses to deviant and omitted speech stimuli with lower amplitudes, longer latencies, and relatively greater frontal topographies than young adults.DiscussionThe results demonstrate that P300 amplitudes elicited in competing noise were a strong predictor of speech-in-noise understanding. This, in addition to the age-related differences in P300 responses, highlights the relevance of neural attentional mechanisms for understanding age-related differences in speech understanding in noise.

In the underwater sound channel we often use an array to receive signals from distant sources. The received signals are often mixed with environmental interference. In the complex acoustic environment, received signals are distorted greatly and elongated in time. In many practical applications such as sound communications, sound remote sensing and active sonar signals, we hope to obtain the original signal's waveform. In general theory, the received signals are the convolution of emission signals and Green's function of environment. In unknown Green's function of environment, simply relying on the array to record the information to determine the sound source signal wave propagation features and the environment is not enough. However, in certain circumstances, based on a physics method of underwater acoustics, the spread of recovery technology is successful.

The human brain has evolved to operate effectively in highly complex acoustic environments, segregating multiple sound sources into perceptually distinct auditory objects. A recent theory seeks to explain this ability by arguing that stream segregation occurs primarily due to the temporal coherence of the neural populations that encode the various features of an individual acoustic source. This theory has received support from both psychoacoustic and functional magnetic resonance imaging (fMRI) studies that use stimuli which model complex acoustic environments. Termed stochastic figure-ground (SFG) stimuli, they are composed of a "figure" and background that overlap in spectrotemporal space, such that the only way to segregate the figure is by computing the coherence of its frequency components over time. Here, we extend these psychoacoustic and fMRI findings by using the greater temporal resolution of electroencephalography to investigate the neural computation of temporal coherence. We present subjects with modified SFG stimuli wherein the temporal coherence of the figure is modulated stochastically over time, which allows us to use linear regression methods to extract a signature of the neural processing of this temporal coherence. We do this under both active and passive listening conditions. Our findings show an early effect of coherence during passive listening, lasting from ∼115 to 185 ms post-stimulus. When subjects are actively listening to the stimuli, these responses are larger and last longer, up to ∼265 ms. These findings provide evidence for early and preattentive neural computations of temporal coherence that are enhanced by active analysis of an auditory scene.

Spectral overlap and temporal avoidance in a tropical savannah frog community