Comparative Analysis of Modality Fusion Approaches for Audio-Visual Person Identification and Verification

Multimodal learning involves integrating information from various modalities to enhance learning and comprehension. We compare three modality fusion strategies in person identification and verification by processing two modalities: voice and face. In this paper, a one-dimensional convolutional neural network is employed for x-vector extraction from voice, while the pre-trained VGGFace2 network and transfer learning are utilized for face modality. In addition, gammatonegram is used as speech representation in engagement with the Darknet19 pre-trained network. The proposed systems are evaluated using the K-fold cross-validation technique on the 118 speakers of the test set of the VoxCeleb2 dataset. The comparative evaluations are done for single-modality and three proposed multimodal strategies in equal situations. Results demonstrate that the feature fusion strategy of gammatonegram and facial features achieves the highest performance, with an accuracy of 98.37% in the person identification task. However, concatenating facial features with the x-vector reaches 0.62% for EER in verification tasks.

Self-Supervised Models in Automatic Whispered Speech Recognition

In automatic speech recognition, any factor that alters the acoustic properties of speech can pose a challenge to the system's performance. This paper presents a novel approach for automatic whispered speech recognition in the Irish dialect using the self-supervised WavLM model. Conventional automatic speech recognition systems often fail to accurately recognise whispered speech due to its distinct acoustic properties and the scarcity of relevant training data. To address this challenge, we utilized a pre-trained WavLM model, fine-tuned with a combination of whispered and normal speech data from the wTIMIT and CHAINS datasets, which include the English language in Singaporean and Irish dialects, respectively. Our baseline evaluation with the OpenAI Whisper model highlighted its limitations, achieving a Word Error Rate (WER) of 18.8% on whispered speech. In contrast, the proposed WavLM-based system significantly improved performance, achieving a WER of 9.22%. These results demonstrate the efficacy of our approach in recognising whispered speech and underscore the importance of tailored acoustic modeling for robust automatic speech recognition systems. This study provides valuable insights into developing effective automatic speech recognition solutions for challenging speech affected by whisper and dialect. The source codes for this paper are freely available.

Deep Neural Networks for Automatic Speaker Recognition Do Not Learn Supra-Segmental Temporal Features

While deep neural networks have shown impressive results in automatic speaker recognition and related tasks, it is dissatisfactory how little is understood about what exactly is responsible for these results. Part of the success has been attributed in prior work to their capability to model supra-segmental temporal information (SST), i.e., learn rhythmic-prosodic characteristics of speech in addition to spectral features. In this paper, we (i) present and apply a novel test to quantify to what extent the performance of state-of-the-art neural networks for speaker recognition can be explained by modeling SST; and (ii) present several means to force respective nets to focus more on SST and evaluate their merits. We find that a variety of CNN- and RNN-based neural network architectures for speaker recognition do not model SST to any sufficient degree, even when forced. The results provide a highly relevant basis for impactful future research into better exploitation of the full speech signal and give insights into the inner workings of such networks, enhancing explainability of deep learning for speech technologies.

Reorganization of the auditory-perceptual space across the human vocal range

We analyzed the auditory-perceptual space across a substantial portion of the human vocal range (220-1046 Hz) using multidimensional scaling analysis of cochlea-scaled spectra from 250-ms vowel segments, initially studied in Friedrichs et al. (2017) J. Acoust. Soc. Am. 142 1025-1033. The dataset comprised the vowels /i y e {\o} {\epsilon} a o u/ (N=240) produced by three native German female speakers, encompassing a broad range of their respective voice frequency ranges. The initial study demonstrated that, during a closed-set identification task involving 21 listeners, the point vowels /i a u/ were significantly recognized at fundamental frequencies (fo) nearing 1 kHz, whereas the recognition of other vowels decreased at higher pitches. Building on these findings, our study revealed systematic spectral shifts associated with vowel height and frontness as fo increased, with a notable clustering around /i a u/ above 523 Hz. These observations underscore the pivotal role of spectral shape in vowel perception, illustrating the reliance on acoustic anchors at higher pitches. Furthermore, this study sheds light on the quantal nature of these vowels and their potential impact on language evolution, offering a plausible explanation for their widespread presence in the world's languages.