JASA express letters最新文献

This Letter investigates the influence of source motion on the performance of the ray-based blind deconvolution algorithm (RBD). RBD is used to estimate channel impulse responses and source signals from opportunistic sources such as shipping vessels but was derived under a stationary source assumption. A theoretical correction for Doppler from a simplified moving source model is used to quantify the biases in estimated arrival angles and travel times from RBD. This correction is numerically validated using environmental data from the SBCeX16 experiment in the Santa Barbara Channel. Implications for source localization and potential passive acoustic tomography using RBD are discussed.

This paper describes an original dataset of children's speech, collected through the use of JIBO, a social robot. The dataset encompasses recordings from 110 children, aged 4-7 years old, who participated in a letter and digit identification task and extended oral discourse tasks requiring explanation skills, totaling 21 h of session data. Spanning a 2-year collection period, this dataset contains a longitudinal component with a subset of participants returning for repeat recordings. The dataset, with session recordings and transcriptions, is publicly available, providing researchers with a valuable resource to advance investigations into child language development.

Variability in speech pronunciation is widely observed across different linguistic backgrounds, which impacts modern automatic speech recognition performance. Here, we evaluate the performance of a self-supervised speech model in phoneme recognition using direct articulatory evidence. Findings indicate significant differences in phoneme recognition, especially in front vowels, between American English and Indian English speakers. To gain a deeper understanding of these differences, we conduct real-time MRI-based articulatory analysis, revealing distinct velar region patterns during the production of specific front vowels. This underscores the need to deepen the scientific understanding of self-supervised speech model variances to advance robust and inclusive speech technology.

Ambient noise measurements from an Arctic fjord during summer and winter are analyzed using spectral, coherence, and directionality estimates from a vertically separated pair of hydrophones. The primary noise sources attributed to wind, shipping, and ice activity are categorized and coherence is arrived at. Estimates of the noise field directionality in the vertical and its variation over time and between seasons are used to strengthen the analysis of the time-varying nature of noise sources. Source identification using such processing techniques serves as a valuable tool in passive acoustic monitoring systems for studying ice dynamics in glacierized fjords.

This study investigates the influence of string properties on bowed string attack playability. To assess the attack playability of different string types, a variety of bow forces and bow accelerations were chosen to excite the strings and measure the transient response under different bowing control parameters. The experimentally obtained playability maps of transient duration as function of bow force and acceleration (Guettler diagram) were obtained with a robotic bowing machine, from four different types of cello G2 strings. Results indicate variations in playability across string types, suggesting that string properties impact attack duration.

In the development of automatic speech recognition systems, achieving human-like performance has been a long-held goal. Recent releases of large spoken language models have claimed to achieve such performance, although direct comparison to humans has been severely limited. The present study tested L1 British English listeners against two automatic speech recognition systems (wav2vec 2.0 and Whisper, base and large sizes) in adverse listening conditions: speech-shaped noise and pub noise, at different signal-to-noise ratios, and recordings produced with or without face masks. Humans maintained the advantage against all systems, except for Whisper large, which outperformed humans in every condition but pub noise.

Far-field (9.7-35.5 km) noise measurements were made during the fifth flight test of SpaceX's Starship Super Heavy, which included the first-ever booster catch. Key results involving launch and flyback sonic boom sound levels include (a) A-weighted sound exposure levels during launch are 18 dB less than predicted at 35 km; (b) the flyback sonic boom exceeds 10 psf at 10 km; and (c) comparing Starship launch noise to Space Launch System and Falcon 9 shows that Starship is substantially louder; the far-field noise produced during a Starship launch is at least ten times that of Falcon 9.

Deep neural network (DNN) based speech enhancement techniques have shown superior performance compared to the traditional speech enhancement approaches in handling nonstationary noise. However, their performance is often compromised as a result of mismatch between their testing and training conditions. In this work, a codebook integrated deep neural network (CI-DNN) approach is introduced for speech enhancement, which mitigates this mismatch by employing existing speaker adapted codebooks with a DNN. The proposed CI-DNN demonstrates better speech enhancement performance compared to the corresponding speaker independent DNNs. The CI-DNN approach essentially involves a post processing operation for DNN and, hence, is applicable to any DNN architecture.

This study investigates the relative perceptual distinction of the [n] vs [l] contrast in different vowel contexts ([_a] vs [_i]) and tonal contexts (high-initial such as HH, HL, vs low-initial such as LL, LH). The results of two speeded AX discrimination experiments indicated that a [n-l] contrast is perceptually more distinct in the [_a] context and with a high-initial tone. The results are consistent with the typology of the [n] vs [l] contrast across Chinese dialects, which is more frequently observed in the [_a] context and with a high-initial tone, supporting a connection between phonological typology and perceptual distinctiveness.

Deriving human neural responses to natural speech is now possible, but the responses to male- and female-uttered speech have been shown to differ. These talker differences may complicate interpretations or restrict experimental designs geared toward more realistic communication scenarios. This study found that when a male talker and a female talker had the same fundamental frequency, auditory brainstem responses (ABRs) were very similar. Those responses became smaller and later with increasing fundamental frequency, as did click ABRs with increasing stimulus rates. Modeled responses suggested that the speech and click ABR differences were reasonably predicted by peripheral and brainstem processing of stimulus acoustics.