The loss of speech function following a laryngectomy usually leads to severe physiological and psychological distress for laryngectomees. In clinical practice, most laryngectomees retain intact upper tract articulatory organs, emphasizing the significance of speech rehabilitation that utilizes articulatory motion information to effectively restore speech. This study proposed a deep learning-based end-to-end method for speech reconstruction using ultrasound tongue images. Initially, ultrasound tongue images and speech data were collected simultaneously with a designed Mandarin corpus. Subsequently, a speech reconstruction model was built based on adversarial neural networks. The model includes a pretrained feature extractor to process ultrasound images, an upsampling block to generate speech, and discriminators to ensure the similarity and fidelity of the reconstructed speech. Finally, both objective and subjective evaluations were conducted for the reconstructed speech. The reconstructed speech demonstrated high intelligibility in both Mandarin phonemes and tones. The character error rate of phonemes in automatic speech recognition was 0.2544, and tone error rate obtained from dictation tests was 0.1784, respectively. Objective results showed high similarity between the reconstructed and ground truth speech. Subjective perception results also indicated an acceptable level of naturalness. The proposed method demonstrates its capability to reconstruct tonal Mandarin speech from ultrasound tongue images. However, future research should concentrate on specific conditions of laryngectomees, aiming to enhance and optimize model performance. This will be achieved by enlarging training datasets, investigating the impact of ultrasound tongue imaging parameters, and further refining this method.

Objective: Extracorporeal shock wave therapy (ESWT) has been widely used in clinical and rehabilitation fields. However, limited knowledge of shock wave sources constrains its therapeutic efficacy. This study aimed to simulate and analyze the characteristics of shock wave sources generated by the radial extracorporeal shock wave therapy (rESWT) device at different media interfaces. Methods: In this paper, a numerical calculation model including the front portion of the handpiece of an rESWT device and three different media (soft tissue, water, and air) was established using the finite element method, and the dynamic impact process was carried out according to the basic principle of the rESWT apparatus. The projectile speed was set to be 7.11, 12.46, 17.81 and 23.16m/s, respectively. In addition, the waveforms at the geometric center of the applicator surface were extracted, and the temporal and spectral characteristics of the shock wave source were disclosed. Results: The waveform of shock wave generated by the applicator of the rESWT device at soft tissue and water interface were similar at the same projectile speed, but both being different from the waveform at air interface. The absolute positive and negative pressure values at the air interface were three orders smaller than those observed at the soft tissue and water interfaces. The spectral characteristics of the waveforms at different media interfaces were similar, with modulation frequencies of soft tissue, water and air being 4.51 kHz, 3.97 kHz and 4.03 kHz, and carrier frequencies being 78.00 kHz, 76.90 kHz and 77.27 kHz, respectively. At different projectile speeds, there was no variation in the waveform shapes of the same medium, revealing that the projectile speed didn't affect the peak frequency of the shock wave waveform. The projectile speed only affected the amplitude of the shock wave waveform, indicating that the absolute positive and negative pressure at the medium interface increased linearly with the projectile speed. Conclusion: The shock waves produced at various media interfaces displayed changes in their temporal and spectral characteristics, and the results observed in biological soft tissues couldn't be directly substituted with those observed in air or water. The findings could offer crucial insights into the assessment of rESWT devices and the enhancement of clinical shock wave treatment protocols.

Extracorporeal shock wave therapy (ESWT), a non-invasive clinical approach, is widely used in sports rehabilitation, orthopaedics, and aesthetic medicine, whose treatment effects result from the interaction of mechanical impulsive waves with target tissues. Though generally accepted, the nature of waves generated by ESWT is not clearly disclosed when propagating through tissues, resulting in the limitation of the technique. In this work, propagation characteristics of shock waves within the tissue-mimicking phantom are revealed. Our approach involves employing a thin, flexible PVDF sensor embedded within the phantoms to capture dynamic wave pressure signals. These signals facilitate the extraction of pertinent features across various impact pressures and depths. Our results demonstrate that the impulsive wave's rise time approximates 5μs. It produces a positive pressure peak ranging from 0.5 to 8.2 MPa, and a negative pressure peak from -0.19 to -3.28 MPa. Notably, peak amplitudes decrease with increasing depth while maintaining a constant impact pressure. As driving pressure increased, maximum positive pressure increased linearly at the same depth. The implications of our findings are pivotal for the design of pertinent devices, formulation of treatment strategies, and assessment of treatment efficacy. Ultimately, a comprehensive understanding of shock wave propagation will enable more informed and effective applications of ESWT.

The electrolarynx (EL) is a common device for voice reconstruction in laryngectomy patients, but its mechanical sound source generates significant radiation noise, affecting the naturalness and acceptability of the speech. The parametric acoustic array (PAA), which produces directionally propagated difference-frequency sound waves, presents a promising alternative for creating a more natural glottal-like voice source in the trachea while reducing radiation noise. In this study, we developed a tissue-mimicking phantom to simulate human neck tissue and used a single-transducer-based PAA platform to generate modulated ultrasound signals with glottal waveform characteristics. Ultrasonic microphones captured sound signals fromthe trachea and surrounding air, and signal processing was used to isolate the difference-frequency signals. The results demonstrated that difference-frequency signals were successfully detected in the phantom’s trachea, with time-domain waveforms and frequency spectra closely resembling the designed glottal waveform (Pearson correlation coefficient = 0.9438). Additionally, radiation noise produced by the PAA was significantly lower (23 dB, p < 0.0001) compared to the traditional EL. These findings demonstrate the potential of PAA for voice source reconstruction in laryngectomy patients and suggest its capacity to enhance speech rehabilitation outcomes. Further research is required to refine the frequency range and evaluate clinical applicability.

Objective: laryngectomees depend on an electromechanical device to generate electrolaryngeal (EL) speech for verbal communication. Compared with normal speech, EL speech suffers from severe distortion, limited phonetic variation, unnatural prosody, and temporal shifts, degrading naturalness and intelligibility. Although sequence-to-sequence (seq2seq) voice conversion (VC) based EL-speech-to-normal-speech conversion (EL2SP) is promising, substantial mismatches between EL and normal speech inevitably cause cumulative mapping errors that limit performance. To address this, we describe a novel representation learning framework integrating speech and text representations to improve mapping and reconstruction quality within a seq2seq VC model. Methods: our methodology comprises two main stages: 1) representation integration and learning, and 2) reconstruction training. A network capable of incorporating auxiliary text information is first constructed with pretrained modules to learn speech–text-based integrated representations. Then, an autoencoder-style reconstruction strategy finalizes EL2SP model to inherit these representations without increasing model complexity. Additional optimization designs are performed across these stages. We introduce three fusion strategies including middle-, input-, and hybrid-level fusion strategies that progressively enhance learning. Moreover, besides standard seq2seq VC objectives, an additional reconstruction loss on the integrated representation is introduced to refine representation transfer. Results: experiments under different EL2SP datasets consistently demonstrate that our methods, combined with data augmentations, outperform baselines relying solely on speech representations regarding both conversion quality and intelligibility. Furthermore, progressive improvements with system design depth validate the effectiveness of our methods. Significance: the proposed methods provide an extensible and practical methodology for EL speech enhancement and assistive communication technologies.