This paper addresses the common challenge of system performance degradation due to speech inconsistency and mismatched acoustic conditions across various domains in speaker verification tasks. We propose a Noise-Aware Quality Network designed to estimate a score based on speech quality and the presence of speech obscured by noise in real-world environments. The score, derived from the normalization of estimated speech quality evaluations, is incorporated into a proposed Noise-Aware Quality loss function, aiming to prioritize speech quality by weighting the embedding distances based on the quality score. Our methodology significantly improves speaker verification performance, particularly in noisy environments. Furthermore, our work highlights the importance of speech quality and the potential benefits of incorporating speech quality weight into the loss function for speaker verification tasks.

Figure 1: System overview of CarAI.Beginning with user-captured photos comprising full-car views for comprehensive inspection coverage and close-up shots for detailed analysis, automatically captured via real-time mobile image quality checking.On the server side, the system identifies key car attributes such as make, model, generation, color, and license plate details.It ensures fraud prevention while also detecting and assessing damages.

In this study, we explore the influence of Thai Linguistic Units (TH-LUs) and speech trimming on the state-of-the-art Thai Text-to-Speech (TTS) systems.We propose an end-to-end Thai TTS framework that emphasizes phonemes, syllables, and words, essential for accurate text pronunciation.To thoroughly investigate these aspects, we designed two main experiments: the TH-LU factor experiment and the TH-LU with speech trimming factor experiment.Our assessment targeted speaker tone and pronunciation accuracy.VITS model demonstrated a standout performer in tonal accuracy, which is evaluated by the Speaker Encoder Cosine Similarity (SECS) method, across different TH-LUs in both trim and non-trim speech training data.For pronunciation accuracy, we integrated a Thai speech-to-text model to evaluate.Our results indicate that VITS with the word linguistic unit outperforms all baselines in overall performance, excelling in both speaker tone and pronunciation accuracy.This research significantly advances the field of TTS, particularly for the Thai language, by highlighting the importance of diverse TH-LU and speech trimming in TTS model development and underlining the need for evaluation methods that account for both tonal accuracy and pronunciation quality.

This paper proposes to adopt concurrent trans- mission in a distributed antenna system to achieve the real- time capability and reliability for wireless automation systems. Our scheme is the simplest way to realize spatial diversity by deploying a large number, e.g. several tens, of transmit antennas and allow all of them to participate in simultaneous packet transmission without any cooperation. We examine the performance, including packet error rate and high- reliability area percentage, in simulation and over-the-air experiments under the IEEE 802.15.4 standard, where the experiments are conducted with up to 32 concurrent transmitters using Tmote Sky sensor nodes. We found that the presence of a timing offset limits the performance. By lowering the timing offset to be below half of a chip period, the reliable performance can be achieved without sacrificing the real-time capability.

Real-time and robust communication is essential for wireless control systems (WCS). Preambles are considered as significant communication overheads in WCS where payloads are typically small. To shorten the overhead in OFDM systems, previous works have considered fixed subcarrier sampling to obtain subcarrier-fixed short preambles (SFSP) and adopted frequency-direction interpolation (FDI) for channel estimation. Despite of advantage in fast fading channel, it results in poor performance in frequency-selective fading channel. In this work, we propose a subcarrier-selectable short preamble (SSSP) which provides full sampling coverage of all subcarriers with several preamble transmissions. By exploiting the periodicity of transmissions in WCS, the SSSP scheme obtains the channel estimation of non-selected subcarriers by not only FDI but also the time-direction extrapolation (TDE), which is robust in frequency-selective fading. In addition, in order to make the channel estimation robust in both fast fading and frequency-selective fading we adopt an adaptive algorithm altering between FDI or TDE mode. The feasibility of the proposed method is validated by evaluating symbol error rate (SER) in fast frequency-selective fading. Simulation results verify that unlike the SFSP the adaptive SSSP scheme shows no error floor in SER curve when the channel is frequency-selective fading, and performs similar to the SFSP scheme when the channel is fast fading.

This paper presents an RSSI-based packet combining (PC) scheme to improve the packet reception reliability affected by a beat interference in wireless sensor networks (WSN) using concurrent transmission (CT). To overcome the beat problem, our PC scheme exploits multiple receptions to perform simple post-detection selection combining based on RSSI values. An implementation verifies its feasibility in resource-limited sensor nodes with time-sensitive CT. Evaluations using computer simulations and hardware experiments show significant reliability improvement.