Soundscape

Augmented Urban Soundscape 

The Smart Nation TRANS Lab has successfully developed a cloud-based, AI-augmented urban soundscape as part of the NRF-funded Cities-of-Tomorrow project. This innovative solution utilizes IoT devices installed in gazebos located in parks adjacent to busy streets. These devices transform the aural environment within the gazebos, making it more pleasant and restoring a sense of calm and tranquility amidst the urban noise. The project also goes beyond perceptual validity by introducing several innovations in soundscape feature extraction, developing perceptual prediction models, and creating an extensive dataset of affective responses augmented by urban soundscapes.

Affective Response Augmented Urban Soundscape (ARAUS)

Choosing optimal maskers for existing soundscapes to achieve desired perceptual changes is challenging due to the variety of maskers and the lack of benchmark datasets. To address this, we have released the ARAUS (Affective Responses to Augmented Urban Soundscapes) dataset, featuring 25,440 unique subjective responses to audio-visual augmented soundscapes with various maskers. We analyzed these responses for consistency and benchmarked the dataset by training and comparing four baseline models for urban soundscape pleasantness.

Autonomous In-Situ Soundscape Augmentation via Joint Selection of Masker and Gain

The selection of maskers and playback gain levels is crucial for enhancing acoustic comfort in soundscape augmentation systems. Traditional methods, relying on expert opinion or listening tests, are often unrepresentative or time-consuming and inflexible to dynamic environments. We developed a deep learning model for joint selection of optimal masker and gain levels, featuring modular components for efficient inference and feature-domain soundscape augmentation, validated by a large-scale study with over 440 participants.

Subjective Perception of Active Noise Reduction and Effect of Informational Masking

Reviving natural ventilation (NV) for urban sustainability poses challenges for indoor acoustic comfort, but active noise control (ANC) and interference-based strategies offer potential solutions. This study examined an active-noise-control (ANC)-based "anti-noise" window (ANW) in a model bedroom, involving 40 participants and various noise types, maskers, and ANC conditions. The results showed that the ANW significantly reduced perceived annoyance and loudness, improved perceived affective quality, and the addition of maskers further enhanced acoustic comfort, although water masking increased loudness.

Lion city soundscapes

This study presents a dataset of audio-visual soundscape recordings at 62 different locations in Singapore, initially made as full-length recordings over spans of 9–38 min. For consistency and reduction in listener fatigue in future subjective studies, one-minute excerpts were cropped from the full-length recordings. An automated method using pre-trained models for Pleasantness and Eventfulness (according to ISO 12913) in a modified partitioning around medoids algorithm was employed to generate the set of excerpts by balancing the need to encompass the perceptual space with uniformity in distribution. A validation study on the method confirmed its adherence to the intended design.

A multidimensional assessment of construction machinery noises based on perceptual attributes and psychoacoustic parameters

Traditional decibel-based measures for predicting annoyance from construction activities are inadequate due to the high acoustic variability of machinery noises. A multidimensional approach using perceptual attributes and psychoacoustic parameters is proposed, with in-situ audio-visual recordings of 16 machinery evaluated for perceived annoyance and perceptual attributes. The study identified three noise clusters based on four perceptual components, correlated with psychoacoustic parameters, to develop an annoyance model that can categorize and manage construction noises effectively.

Quality assessment of acoustic environment reproduction methods for cinematic virtual reality in soundscape applications

Spatial audio, crucial in virtual reality (VR) research, was assessed using three methods—FOA-static binaural, FOA-tracked binaural, and FOA 2D octagonal speaker array—down-mixed from first-order ambisonics (FOA). In a study involving 30 participants, subjective evaluations compared in situ and VR representations across four outdoor locations. While overall soundscape quality showed no significant differences, spatial qualities varied significantly across reproduction methods, with FOA-tracked binaural and FOA 2D speaker array methods demonstrating superior spatial fidelity. Specifically, the FOA-tracked binaural method was highlighted for its adequacy in evaluating cinematic VR soundscapes.

Effects of Adding Natural Sounds to Urban Noises on the Perceived Loudness of Noise and Sopuncdscape Quality

Introducing pleasant natural sounds to mask urban noises is a crucial strategy for improving acoustic comfort. This study examines the effects of signal-to-noise ratio (SNR) and temporal characteristics of natural sounds (birdsong and stream) on perceived loudness of noise (PLN) and overall soundscape quality (OSQ) through experiments with hydraulic breaker and traffic noises at various sound pressure levels (SPL). Results show that adding natural sounds reduced PLN by 17.9% and increased OSQ by 18.3%, with optimal SNRs identified as -6 dB when noise levels reached 75 dB.