The sound-source localization provided by a crosstalk cancellation (CTC) system depends on the head-related transfer functions (HRTFs) used for the CTC filter calculation. In this study, the horizontal- and sagittal-plane localization performance was investigated in humans listening to individualized matched, individualized but mismatched, and non-individualized CTC systems. The systems were simulated via headphones in a binaural virtual environment with two virtual loudspeakers spatialized in front of the listener. The individualized mismatched system was based on two different sets of listener-individual HRTFs. Both sets provided similar binaural localization performance in terms of quadrant, polar, and lateral errors. The individualized matched systems provided performance similar to that from the binaural listening. For the individualized mismatched systems, the performance deteriorated, and for the non-individualized mismatched systems (based on HRTFs from other listeners), the performance deteriorated even more. The direction-dependent analysis showed that mismatch and lack of individualization yielded a substantially degraded performance for targets placed outside of the loudspeaker span and behind the listeners, showing relevance of individualized CTC systems for those targets. Further, channel separation was calculated for different frequency ranges and is discussed in the light of its use as a predictor for the localization performance provided by a CTC system.

The auditory brain circuits are biologically constructed to recognize and localize sounds by encoding a combination of cues that help individuals interpret sounds. The development of computational methods inspired by human capacities has established opportunities for improving machine hearing. Recent studies based on deep learning show that using convolutional recurrent neural networks (CRNNs) is a promising approach for sound event detection and localization in spatial sound. Nevertheless, depending on the sound environment, the performance of these systems is still far from reaching perfect metrics. Therefore, this work intends to boost the performance of state-of-the-art (SOTA) systems by using bio-inspired gammatone auditory filters and intensity vectors (IVs) for the acoustic feature extraction stage, along with the implementation of a temporal convolutional network (TCN) block into a CRNN model, to capture long term dependencies. Three data augmentation techniques are applied to increase the small number of samples in spatial audio datasets. The mentioned stages constitute our proposed Gammatone-based Sound Events Localization and Detection (G-SELD) system, which exceeded the SOTA results on four spatial audio datasets with different levels of acoustical complexity and with up to three sound sources overlapping in time.

Dynamic crosstalk cancellation (CTC) systems commonly find use in immersive virtual reality (VR) applications. Such dynamic setups require extremely high filter update rates, so filter calculation is usually performed in the frequency-domain for higher efficiency. This paper proposes a general framework for the calculation of dynamic CTC filters to be used in immersive VR applications. Within this framework, we introduce a causality constraint to the frequency-domain calculation to avoid undesirable wrap-around effects and echo artifacts. Furthermore, when regularization is applied to the CTC filter calculation, in order to limit the output levels at the loudspeakers, noncausal artifacts appear at the CTC filters and the resulting ear signals. We propose a global minimum-phase regularization to convert these anti-causal ringing artifacts into causal artifacts. Finally, an aspect that is especially critical for dynamic CTC systems is the filter switch between active loudspeakers distributed in a surround audio-visual display system with 360 ^° of freedom of operator orientation. Within this framework we apply a weighted filter calculation to control the filter switch, which allows the loudspeakers' contribution to be windowed in space, resulting in a smooth filter transition.

In this study, we examined the effects of project-based learning (PBL) on student learning outcomes related to the subject of signals and systems in the field of electronic engineering at the Universidade Estadual de Campinas (UNICAMP) in Brazil and at the Pontificia Universidad Javeriana (PUJ) in Colombia. We used two methods to assess the effect of PBL on student outcomes: (1) we used the Signals and Systems Concept Inventory (SSCI) to measure the increase in conceptual understanding of signals and systems among electronic engineering UNICAMP students as a consequence of implementing PBL; and (2) we compared the results on a comprehensive signals and systems final exam of a group of electronic engineering students at PUJ who received PBL to those who did not. Results indicated that (1) UNICAMP students achieved outcomes comparable to those of Buck and Wage’s study: UNICAMP students taught with projects learned more than students in 15 Signal and Systems lecture-based courses in the United States; and (2) PUJ students taught with projects received higher final exam grades than students taught via lectures. Students were able to apply their knowledge of signal processing and systems analysis using MATLAB models. These models provide authentic assessments of engineering students’ knowledge and skills. The findings of this study indicate that PBL is more effective than lectures in enhancing students’ understanding and application of signals and systems concepts.

Sound Event Localization and Detection (SELD) is a critical challenge in various industrial applications, such as autonomous systems, smart cities, and audio surveillance, which require accurate identification and localization of sound events in complex environments. Traditional supervised approaches heavily rely on large, annotated multichannel audio datasets, which are expensive and time-consuming to produce. This paper addresses this limitation by introducing the w2v-SELD architecture, a self-supervised model adapted from the wav2vec 2.0 framework to learn effective sound event representations directly from raw, unlabeled 3D audio data. The proposed model follows a two-stage process: pre-training on large, unlabeled 3D audio datasets to capture high-level features, followed by fine-tuning with a smaller, labeled SELD dataset. Experimental results show that our w2v-SELD method outperforms baseline models on Detection and Classification of Acoustic Scenes and Events (DCASE) challenges, achieving a 66% improvement for DCASE TAU-2019 and a 57% improvement on DCASE TAU-2020 with respect to baseline systems. The w2v-SELD model performs competitively with state-of-the-art supervised methods, highlighting its potential to significantly reduce the dependency on labeled data in industrial SELD applications. The code and pre-trained parameters of our w2v-SELD model are available in this repository.