Perceptual Evaluation Research Articles

A systematic review of deep learning-based denoising for low-dose computed tomography from a perceptual quality perspective.

Low-dose computed tomography (LDCT) scans are essential in reducing radiation exposure but often suffer from significant image noise that can impair diagnostic accuracy. While deep learning approaches have enhanced LDCT denoising capabilities, the predominant reliance on objective metrics like PSNR and SSIM has resulted in over-smoothed images that lack critical detail. This paper explores advanced deep learning methods tailored specifically to improve perceptual quality in LDCT images, focusing on generating diagnostic-quality images preferred in clinical practice. We review and compare current methodologies, including perceptual loss functions and generative adversarial networks, addressing the significant limitations of current benchmarks and the subjective nature of perceptual quality evaluation. Through a systematic analysis, this study underscores the urgent need for developing methods that balance both perceptual and diagnostic quality, proposing new directions for future research in the field.

A novel autonomous exploration algorithm via LiDAR/IMU SLAM and hierarchical subsystem for mobile robot in unknown indoor environments

Abstract Autonomous exploration in unknown environments is an essential capability for mobile robots. The complexity of autonomous exploration, however, means that existing algorithms struggle to balance efficiency and comprehensiveness, causing low mapping accuracy and redundant path planning. To perform accurate and efficient exploration tasks, we have proposed a novel autonomous exploration algorithm via LiDAR/IMU (Inertial Measurement Unit) Simultaneous Localization and Mapping (SLAM) and hierarchical subsystem for mobile robot in unknown environments. Firstly, to enhance mapping accuracy for mobile robot exploration, LiDAR/IMU SLAM is improved with the assistance of backward propagation and iterated Kalman filter, and Bidirectional Rapidly–Exploring Random Trees* (BI–RRT*) is applied for efficient frontier point detection. Secondly, we optimize local path planning by leveraging information theory through perceptual quality evaluation, which is then integrated with global path planning utilizing an enhanced Travelling Salesman Problem solver and a sparse grid map to amplify exploration efficiency. Thirdly, an enhanced hierarchical autonomous exploration method for mobile robots is proposed, which incorporates local path planning for seamless navigation around highly promising exploration spots, coupled with global path planning to effectively interconnect various sub–regions. Finally, simulations and field tests have demonstrated that the proposed method explores an unknown indoor environment with a 30.8% reduction in exploration time and a 29.9% reduction in exploration path in comparison with Dynamic Stage Viewpoint Planner. The map constructed in this paper has more accurate details and exploration paths have been shortened to ensure effective exploration.

Enhanced Speech Processing for Air Traffic Control Using an Optimized Generative Adversarial Network

In order to improve the quality of voice calls during air traffic control, an improved SEGAN air traffic control speech enhancement algorithm is proposed. Aiming at the problem that the traditional speech enhancement algorithm based on generative adversarial network (SEGAN) is drowned under low signal-to-noise ratio conditions, a multi-stage, multi-mapping, multi-dimensional output generator and a multi-scale, multi-discriminator network model are proposed based on the SEGAN network model. First, the speech semantic features are extracted based on the deep neural network structure to complete the semantic segmentation of air traffic control speech. Secondly, multiple generators are set to further optimize the speech signal. Then, a downsampling module is added to the convolution layer to improve the model's utilization of speech information and reduce the loss of speech information. Finally, multi-scale, multiple discriminators are used to learn the distribution law and information of speech samples in multiple directions. The results show that under low signal-to-noise ratio conditions, the improved SEGAN model improves the short-term objective intelligibility (STOI) and the perceptual evaluation of speech quality (PESQ) by and respectively , which can quickly and effectively perform air traffic control speech enhancement and provide preparation for subsequent air traffic control speech recognition.

Perceived Quality of Binaural Rendering from Baffled Microphone Arrays Evaluated Without an Explicit Reference

The authors present a perceptual evaluation of the binaural rendering quality of signals from several types of baffled microphone arrays. They employ the multi-stimulus category rating (MuSCR) paradigm that does not require a reference stimulus. The tested conditions also comprise a very high numerical accuracy stimulus, given the highest quality rating in approximately half of the multi-stimulus trials. A comparison with the literature on spherical microphone arrays (SMAs) shows that MuSCR allows for drawing representative conclusions regarding the dependency of the perceived quality on the array and rendering parameters as in previous experiments with an explicit high-fidelity reference. The authors applied the MuSCR paradigm to evaluate the perceived reproduction quality of equatorial microphone arrays (EMAs) with microphones only along the equator of the spherical baffle and of equatorial arrays with a nonspherical baffle (XMAs). The results endorse the observations from SMAs that an increasing spherical harmonic order leads to improved perceived quality. The authors also confirm that EMAs lead to the same perceived quality as SMAs despite the substantial difference in the number of microphones. Magnitude equalization of artifacts from spatial undersampling can be very effective for XMAs the raw solution deviates significantly from the other array types at high frequencies.

Impact of Mask Type as Training Target for Speech Intelligibility and Quality in Cochlear-Implant Noise Reduction

The selection of a target when training deep neural networks for speech enhancement is an important consideration. Different masks have been shown to exhibit different performance characteristics depending on the application and the conditions. This paper presents a comprehensive comparison of several different masks for noise reduction in cochlear implants. The study incorporated three well-known masks, namely the Ideal Binary Mask (IBM), Ideal Ratio Mask (IRM) and the Fast Fourier Transform Mask (FFTM), as well as two newly proposed masks, based on existing masks, called the Quantized Mask (QM) and the Phase-Sensitive plus Ideal Ratio Mask (PSM+). These five masks are used to train networks to estimate masks for the purpose of separating speech from noisy mixtures. A vocoder was used to simulate the behavior of a cochlear implant. Short-time Objective Intelligibility (STOI) and Perceptual Evaluation of Speech Quality (PESQ) scores indicate that the two new masks proposed in this study (QM and PSM+) perform best for normal speech intelligibility and quality in the presence of stationary and non-stationary noise over a range of signal-to-noise ratios (SNRs). The Normalized Covariance Measure (NCM) and similarity scores indicate that they also perform best for speech intelligibility/gauging the similarity of vocoded speech. The Quantized Mask performs better than the Ideal Binary Mask due to its better resolution as it approximates the Wiener Gain Function. The PSM+ performs better than the three existing benchmark masks (IBM, IRM, and FFTM) as it incorporates both magnitude and phase information.

Reactive Gaze during Locomotion in Natural Environments

AbstractAnimating gaze behavior is crucial for creating believable virtual characters, providing insights into their perception and interaction with the environment. In this paper, we present an efficient yet natural‐looking gaze animation model applicable to real‐time walking characters exploring natural environments. We address the challenge of dynamic gaze adaptation by combining findings from neuroscience with a data‐driven saliency model. Specifically, our model determines gaze focus by considering the character's locomotion, environment stimuli, and terrain conditions. Our model is compatible with both automatic navigation through pre‐defined character trajectories and user‐guided interactive locomotion, and can be configured according to the desired degree of visual exploration of the environment. Our perceptual evaluation shows that our solution significantly improves the state‐of‐the‐art saliency‐based gaze animation with respect to the character's apparent awareness of the environment, the naturalness of the motion, and the elements to which it pays attention.

Tongue and Lip Strength and Endurance in Healthy Young Adults With and Without Orofacial MyofunctionalDisorders.

Based on the premise that tongue and lip force is affected in individuals with orofacial myofunctional disorders (OMDs), orofacial myofunctional therapy (OMT) programs typically include the strengthening of orofacial muscles through isometric and isotonic exercises. The purpose of this study is to examine whether there is indeed a measurable difference in maximum tongue and lip strength and endurance between subjects with and without OMDs, as well as to explore Oral Health-Related Quality of Life (OHRQoL) in this population. Sixty healthy young adults (mean age 18.9 years, SD 0.69 years, range 18.0-21.4 years) participated in this study. Perceptual evaluation of orofacial functions was conducted using the Orofacial Myofunctional Evaluation with Scores (OMES) protocol. Participants were divided into three groups, namely those with physiological orofacial functions (POF; n = 20), those with inconsistent orofacial functions (IOF; n = 11), and those with consistent OMDs (OMD; n = 29). Maximum isometric pressure and endurance of tongue and lips were measured using the Iowa Oral Performance Instrument (IOPI). OHRQoL was assessed using the Oral Health Impact Profile (OHIP-14). The OMES total scores did not differ significantly between the three groups, though the subscale 'function' showed significant differences (13.637; p = 0.001). No statistically significant differences in maximum isometric pressure or endurance for tongue and lips, nor for OHIP scores, were measured between the three groups. Although orofacial muscle strengthening is commonly part of OMT, current findings suggest the superfluity of these types of exercises in the treatment of OMDs in healthy young adults and may indicate the need to focus on awareness and rehabilitation of pathological functions in these individuals. Re-evaluation of the efficacy of OMT programs in different populations may be an important strategy going forward.

Speech Outcome Audit for Cleft Lip and Palate Patient Population Correlated With Veau Cleft Type, Surgical Repair Technique, and Patient Age

Objective: The objective of this study is to determine speech outcome differences for specific palate repair techniques and correlate these data with patient age at the time of operation. Methods: A retrospective study was conducted on a cohort of consecutive nonsyndromic patients who underwent cleft lip and/or palate repair at the authors' hospital between 2010 and 2020. Only those patients who participated in at least 4 years of follow-up accompanied by audio-video recording were included. The patient cohort was divided into 2 age groups, between 1 year and 2 years of age versus older than 2 years of age. Standardized perceptual speech evaluations and nasopharyngoscopy were performed by the authors' team to detect the presence of velopharyngeal insufficiency. Results: A total of 481 patients were initially evaluated, and 323 of these patients met all inclusion criteria and were subsequently selected for this study. One hundred seventy-two patients (n = 172) underwent palatal surgery using the Bardach 2-flap palatoplasty technique, 47 patients (n = 47) underwent repair through the von Langenbeck palatoplasty technique, 78 patients (n = 78) underwent repair through Veau-Wardill-Killner palatoplasty technique, and 26 patients (n = 26) underwent repair through double-opposing Z-plasty palatoplasty. The rate of residual velopharyngeal insufficiency was statistically similar among all palatoplasty techniques. A comparison between these two age groups revealed that patients younger than 2 years of age showed higher speech outcome success. Conclusion: High rates of speech success are achieved for patients younger than 2 years of age. Speech outcomes success is maximized by proper technique selection in relation to cleft type.

Optimisation strategies for indoor acoustic environment in Elderly apartments from a multi-sensory perspective

As ageing intensifies, the perceptual evaluation of the elderly is increasingly valued. Research has shown that appropriate soundscape can have a positive effect on environmental perception, but current soundscape design in elderly apartments pays more attention to auditory perception and often neglects the comprehensive impact of multi-sensory. This study was conducted in two elderly apartments in Wuxi and Jinan through physical environment measurements, comparative experiments and questionnaires. The study tested the indoor physical environment and collected the elderly's perceptual evaluation of each physical environment index. Combined with acoustic and olfactory environment comparative experiments, the study proposed strategies for optimising indoor acoustic and olfactory environment to meet the comfort level of the Chinese elderly. The participants were aged 70-90 years old and about 100 questionnaires were obtained. The results of the study found that the acoustic environment of elderly apartments, which is mainly based on natural sound, better enhances elderly's perceptual evaluation of the indoor environment; there are differences in the physical environment needs of older people for different functional rooms.; comfortable odours significantly enhance elderly's overall evaluation of the environment; and different physical environment indexes interact with elderly's perception.

Perceptual evaluation of sound synthesis of wind turbine noise

This study brings together several laboratories with the aim of assessing the health effects of "audible" noise (> 20 Hz) and infrasound (< 20 Hz) emitted by wind turbines. To study loudness and annoyance due to this noise, perceptual tests are planned at the LMA, where a restitution cabin has been developed, specifically designed to diffuse very low frequencies and infrasound. As recording wind turbine noise is only possible at low wind speeds for a good quality sound reproduction, it would be interesting to be able to use sound synthesis of wind turbine noise. From sounds recorded in a wind farm for different meteorological conditions, the corresponding sounds have been synthesized. A physical model synthesis was performed, based on an extended-source aeroacoustic model taking into account propagation over flat ground. Dissimilarity tests including recorded and synthesized sounds enabled a 2D perceptual space to be built. Synthesized and the corresponding recorded sounds are closed together in the perceptual space, but some differences can be perceived, mainly due to difference in amplitude of fluctuation and spectral balance. The analysis of the perceptual space opens up interesting prospects for improving the sound synthesis and its use for future perceptual tests.

Realization of global audio telepresence via a learning-based model-matching approach with an acoustic array system

A Global Audio Telepresence (GOAT) system requires a microphone array to capture the spatial audio signals in the far end and a loudspeaker array to reconstruct the sound field in the near end. This seamlessly immerses near-end users in remote audio scenes with full ambience. In this paper, we use a learning-based GOAT system (L-GOAT) based on the model-matching principle, where a deep neural network (DNN) acts as non-linear filters for the GOAT system. The network training attempts to minimize the matching error between the signals reproduced by the DNN and the desired signals filtered by the far-end acoustic transfer functions (ATFs). Extensive simulations were carried out for multi-source scenarios in two different rooms with different reverberation times. To implement the L-GOAT system, a five-microphone linear array was adopted in the far-end room, while a six-loudspeaker array was utilized in the near-end room. The objective evaluation matrices, including the Perceptual Evaluation of Speech Quality (PESQ), Short-Time Objective Intelligibility (STOI), and the matching errors, were conducted to validate the efficacy of the GOAT systems. The proposed learning-based approach has demonstrated superior performance compared to a conventional digital signal processing (DSP)-based method.

A Comparative Effectiveness Study of Speech and Surgical Outcomes: Study Overview.

To provide an overview of the Cleft Outcomes Research NETwork (CORNET) and the CORNET Speech and Surgery study. The study is (1) comparing speech outcomes and fistula rate between two common palate repair techniques, straight-line closure with intra-velar veloplasty (IVVP) and Furlow Double-Opposing Z-palatoplasty (Furlow Z-plasty); (2) summarizing practice variation in the utilization of early intervention speech-language (EI-SL) services; and (3) exploring the association between EI-SL services and speech outcomes. Prospective, longitudinal, observational, comparative effectiveness, multi-center. Twenty sites across the United States. One thousand two hundred forty-seven children with cleft palate with or without cleft lip (CP ± L). Children with submucous cleft palate or bilateral sensorineural severe to profound hearing loss were excluded from participation. Straight-line closure with IVVP or Furlow Z-plasty based on each surgeon's standard clinical protocol. The primary study outcome is perceptual ratings of hypernasality judged from speech samples collected at 3 years of age. Secondary outcomes are fistula rate, measures of speech production, and quality of life. The statistical analyses will include generalized estimating equations with propensity score weighting to address potential confounders. Recruitment was completed in February 2023; 80% of children have been retained to date. Five hundred sixty two children have completed their final 3-year speech assessment. Final study activities will end in early 2025. This study addresses long-standing questions related to the effectiveness of the two most common palatoplasty approaches and describes CORNET which provides an infrastructure that will streamline future studies in all areas of cleft care.

Adaptation and evaluation of Pyroomacoustics for auralization purposes

An open-source room simulation software that relies on geometrical acoustics - Pyroomacoustics - was used to compute Binaural Room Impulse Responses (BRIRs) of existing rooms that are modeled in the Benchmark for Room Acoustical Simulation (BRAS). Modifications were made to the software to investigate the impact of the scattering model on the acoustic parameter values. A comparison with the acoustic parameter measurements showed that the values of EDT, C80 and D50 were found to be closer when using the Vector-Mixing method in comparison with the original scattering method. Overall, the alternative method reduced by 95% the mean squared error on the acoustic parameter values. The sound source directivity was also considered in the simulations to quantify its influence on the perceived auditory scenes. A listening test was then conducted to compare binaural renderings derived from measured and simulated BRIRs. Results showed that ratings of the modified versions of the software were either not significantly different or significantly higher than those of the original version depending on the room and sound source. Including the source directivity significantly improved the perceptual ratings, which is thought to be due to the reduction of spectral and loudness differences with the measured reference.

Multimodal Hong Kong: a multisensory study of environmental perception of places through sight, sound, and smell

Despite the importance of soundscape and smellscape in intangible cultural heritage, little is known about multisensory interactions. To explore the impact of multiple senses on environmental perception, the present study focused on six stalls at Shek Kip Mei wet market in Hong Kong. In the first part of the study, an online survey collected perceptual ratings of whole, auditory, and visual environmental quality, as well as free-form responses of imaginations of smell. Participants (N = 36) were presented with 12 audiovisual captures where audio was either matched with video or mismatched. ANOVA was used to examine the influence of three independent factors - scenes, modes, and a priori valence groups - on the ratings. The imagined smells were analysed through frequency counts of keywords. The results demonstrate participants' capacity to notice changes in audio modes and to perceive the environment through the recognition of scene context and multisensory interactions. In the second part of the study, not discussed in the present paper, we will extend this experimental method to include smells in a laboratory setting. We discuss the importance of understanding multisensory interactions between sight, sound, and smell for environmental perceptions and Hong Kong's intangible cultural heritage and urban studies.

Effect of Modulation Domain Coupled KalmanSpectral Filter on Speech Enhancement over Wireless Voiced Communication System

Noise suppression in speech signal has received great attention in signal processing research community. However, existing methods of noise suppression such as the time-domain Kalman filter suffers from the psychoacoustic and physiological characteristic feature while the spectral subtraction technique performed in the modulation domain is associated with remnant noise in the enhanced speech signal. Therefore, in this paper, a Modulation-Domain coupled Kalman-Spectral Filtering (MD-KSF) in a single channel system is proposed to address the aforementioned shortcomings in the two existing techniques. The simulation of the proposed coupled MD-KSF was performed in MATLAB software environment. The evaluation of the proposed coupled MD-KSF technique in terms of Spectral waveform, Mean Square Error (MSE) and Log Spectral Distance (LSD) was performed. Furthermore, Perceptual Evaluation of Speech Quality (PESQ) and Short-Time Objective Intelligibility (STOI) tests were employed to validate the quality and intelligibility of the proposed coupled MD- KSF using the NOIZEUS corpus data set. The proposed technique shows significant noise suppression over the existing techniques.

Evaluation of the effect of endoscopic CO₂laser arytenoidectomy on re-operation for failed posterior cordotomy

Objective:To analyze the efficacy of endoscopic CO₂ laser arytenoidectomy in treating bilateral vocal cord paralysis. Methods:Ninety-five patients who underwent endoscopic CO₂ laser arytenoidectomy for bilateral vocal cord paralysis at the First Affiliated Hospital of Naval Medical University（Changhai Hospital） of Shanghai from January 2009 to December 2022 were included in this study. Among them, 59 patients underwent endoscopic CO₂ laser arytenoidectomy as their first glottic enlargement surgery（Group A）, while 36 patients, who had previously undergone two unsuccessful CO₂ laser-assisted posterior cordotomies, underwent endoscopic CO₂ laser arytenoidectomy as a subsequent surgery（Group B）. Swallowing function, electronic laryngoscopy, perceptual voice evaluation, and objective voice analysis indicators were statistically analyzed before and after surgery to evaluate clinical efficacy. Results:The extubation rate after the first surgery was 84.75% in Group A and 86.11% in Group B, with total extubation rates of 94.92% and 94.44%, respectively. There were no significant differences between the two groups in preoperative and postoperative swallowing function, glottic size, or various voice evaluation indicators（P>0.05）. Within-group comparisons showed that postoperative swallowing function, glottic closure during phonation, perceptual evaluations of G（grade of hoarseness）, A（asthenia）, and B（breathiness） significantly worsened, with increased grades. The maximum transverse diameter of the posterior glottis during inspiration significantly increased, and the VHI-10 score was significantly higher postoperatively. Jitter, shimmer, and the harmonics-to-noise ratio significantly deteriorated, and maximum phonation time significantly shortened（P<0.05）. No significant differences were observed in postoperative R（roughness） and S（strain） compared to preoperative values（P>0.05）. Conclusion:Endoscopic CO₂ laser arytenoidectomy can impair voice quality to some extent but effectively alleviates breathing difficulties in patients with bilateral vocal cord paralysis. For patients who did not achieve successful extubation with CO₂ laser-assisted posterior cordotomy, endoscopic CO₂laser arytenoidectomy is an effective reoperative method, ensuring a high extubation rate while preserving certain voice functions.

Voice analysis in women with polycystic ovary syndrome

BackgroundPolycystic ovarian syndrome (PCOS) is a common endocrine disorder affecting one in five women of reproductive age. Polycystic appearance of the ovaries in ultrasonography is associated with oligomenorrhea, and clinical and or biochemical hyperandrogenism. PCOS has been associated with voice alterations for many years. However, there is insufficient research on voice analysis in PCOS patients.ObjectivesThis research aimed to examine the voice characteristics among women with PCOS.Patients and methodsOur study included 25 women confirmed with PCOS and 25 control. A complete voice assessment was performed for all participants using primary demographic data, vocal symptoms, auditory perceptual evaluation of the voice, Arabic version of the Voice Handicap Index, acoustic analysis, and video laryngoscopic examination of the larynx.ResultsOur study revealed statistically significant differences between PCOS and control groups in phonasthenic symptoms, auditory perceptual assessment of the voice, acoustic parameters, and laryngoscopic examination.ConclusionPCOS women with hirsutism are complaining of vocal symptoms, changes in vocal characters, and deterioration of the acoustic parameters of the voice. High serum levels of androgen can cause the deepening of the voice, which is essential for the early detection of vocal symptoms in PCOS patients to prevent virilization of their voice.Trial registrationNCT06085170. Registered 20 August 2023—retrospectively registered, https://www.google.com/url?sa=t&source=web&rct=j&opi=89978449&url=https://classic.clinicaltrials.gov/ct2/show/NCT06085170&ved=2ahUKEwjEyLqUq6KGAxWxRPEDHb57DIQQFnoECBQQAQ&usg=AOvVaw1xTurnA6mJEZ9PK0vQS8VL.

A generative-AI-based design methodology for car frontal forms design

With the advancement of artificial intelligence, big data, and cloud computing, numerous generative AI applications have surfaced. In contrast to conventional generative design and computer-aided design tools, these applications significantly enhance design productivity. However, there are currently few design methodologies based on generative AIs in the academic community to improve the efficiency of industrial designers and optimize the design process. This study introduces a creative and practical methodology for designing car frontal forms based on generative AIs. In this methodology, imagery adjectives describing car frontal forms are generated by sending valid prompts to the text-generative AI application GPT-4.0. Then input typical imagery adjectives into the image-generative AI Midjourney successively as prompts to generate many car frontal forms that align with the typical imagery adjectives, forming a reference form database. Subsequently, a base form is selected, and target imageries are defined. Simultaneously, forms from the reference form database, conforming to the target imageries, are chosen as the reference forms. The main form elements of the base and reference forms are then delineated using cubic Bézier curves. Finally, a form curve blending algorithm is applied to obtain a set of alternatives, and the image-generative AI application Vega AI is utilized to convert the alternatives into three-dimensional renderings. FAHP-based expert evaluation and consumer perceptual evaluation are employed to validate the alternatives. Results indicate that the alternatives effectively capture the imageries of the reference forms. The proposed generative-AI-based design methodology enhances the efficiency of industrial designers, thereby minimizing human and material costs in product development. Additionally, this study presents a design case using various generative AIs to inspire designers to re-examine the traditional design process.

Decode, Move and Speak! Self-supervised Learning of Speech Units, Gestures, and Sound Relationships Using Vocal Imitation

Abstract Speech learning encompasses mastering a complex motor system to produce speech sounds from articulatory gestures while simultaneously uncovering discrete units that provide entry to the linguistic system. Remarkably, children acquire these associations between speech sounds, articulatory gestures, and linguistic units in a weakly supervised manner, without the need for explicit labeling of auditory inputs or access to target articulatory gestures. This study uses self-supervised deep learning to investigate the respective roles of sounds, gestures, and linguistic units in speech acquisition and control. In a first experiment, we analyzed the quantized representations learned by vector-quantized variational autoencoders (VQ-VAE) from ground truth acoustic and articulatory data using ABX tests. We show an interesting complementarity between acoustic and articulatory modalities that may help in the discovery of phonemes. In a second experiment, we introduce a computational agent that repeats auditory speech inputs by controlling a virtual vocal apparatus. This agent integrates an articulatory synthesizer capable of reproducing diverse speech stimuli from interpretable parameters, along with two internal models implementing the articulatory-to-acoustic (forward) and acoustic-to-articulatory (inverse) mapping, respectively. Additionally, two inductive biases are used to regularize the ill-posed acoustic-to-articulatory inverse mapping. In line with the first experiment, we explore the complementarity between the auditory input and the articulatory parameters inferred by the agent. We also evaluate the impact of discretizing auditory inputs using VQ-VAE. While the majority of the agent’s productions are intelligible (according to perceptual evaluations), our analysis highlights inconsistencies in the underlying articulatory trajectories. In particular, we show that the agent’s productions only partially reproduce the complementarity between the auditory and articulatory modalities observed in humans.

Artificial Intelligence-Assisted Speech Therapy for /ɹ/: A Single-Case Experimental Study.

This feasibility trial describes changes in rhotic production in residual speech sound disorder following ten 40-min sessions including artificial intelligence (AI)-assisted motor-based intervention with ChainingAI, a version of Speech Motor Chaining that predicts clinician perceptual judgment using the PERCEPT-R Classifier (Perceptual Error Rating for the Clinical Evaluation of Phonetic Targets). The primary purpose is to evaluate /ɹ/ productions directly after practice with ChainingAI versus directly before ChainingAI and to evaluate how the overall AI-assisted treatment package may lead to perceptual improvement in /ɹ/ productions compared to a no-treatment baseline phase. Five participants ages 10;7-19;3 (years;months) who were stimulable for /ɹ/ participated in a multiple (no-treatment)-baseline ABA single-case experiment. Prepractice activities were led by a human clinician, and drill-based motor learning practice was automated by ChainingAI. Study outcomes were derived from masked expert listener perceptual ratings of /ɹ/ from treated and untreated utterances recorded during baseline, treatment, and posttreatment sessions. Listeners perceived significantly more rhoticity in practiced utterances after 30 min of ChainingAI, without a clinician, than directly before ChainingAI. Three of five participants showed significant generalization of /ɹ/ to untreated words during the treatment phase compared to the no-treatment baseline. All five participants demonstrated statistically significant generalization of /ɹ/ to untreated words from pretreatment to posttreatment. PERCEPT-clinician rater agreement (i.e., F1 score) was largely within the range of human-human agreement for four of five participants. Survey data indicated that parents and participants felt hybrid computerized-clinician service delivery could facilitate at-home practice. This study provides evidence of participant improvement for /ɹ/ in untreated words in response to an AI-assisted treatment package. The continued development of AI-assisted treatments may someday mitigate barriers precluding access to sufficiently intense speech therapy for individuals with speech sound disorders. https://doi.org/10.23641/asha.26662807.