arXiv subjects

Aref Farhadipour

Publications and source records attributed to Aref Farhadipour.

15 recordsLinked to original sources

arXiv2026

GRAFT: Grafted Reference Audio for Fine-grained Pronunciation in Zero-shot Text-to-Speech

We present GRAFT, a per-word pronunciation conditioning mechanism for text-to-speech neural codec language modeling. Existing systems reach high intelligibility and naturalness but inherit the ambiguity of text and mispronounce rare proper nouns, loanwords and technical terms. Even phoneme-conditioned models offer no direct acoustic handle for per-word pronunciation. GRAFT controls the pronunciation of a chosen word from a short spoken sample of it, encoded with the model's own speech tokenizer and bound to the word's position in the prompt. Voice conversion during training-data construction disentangles the hint speaker from the target speaker, so the hint may come from any voice while the output stays in the target voice. In a blind English listening study, human raters rank GRAFT first by a clear margin, judging its rendering of the difficult word closest to a reference recording of that word. On a five-language objective benchmark, GRAFT reduces target-word phoneme error rate by 22-39% over the identical text-only backbone and outperforms competitive open-source zero-shot systems, both phoneme- and text-conditioned, on target-word pronunciation, while preserving speaker similarity and naturalness.

cs.LG↗

arXiv2026

Foundation Model Embeddings Meet Blended Emotions: A Multimodal Fusion Approach for the BLEMORE Challenge

We present our system for the BLEMORE Challenge at FG 2026 on blended emotion recognition with relative salience prediction. Our approach combines six encoder families through late probability fusion: an S4D-ViTMoE face encoder adapted with soft-label KL training, frozen layer-selective Wav2Vec2 audio features, finetuned body-language encoders (TimeSformer, VideoMAE), and -- for the first time in emotion recognition -- Gemini Embedding 2.0, a large multimodal model whose video embeddings produce competitive presence accuracy (ACCP = 0.320) from only 2 seconds of input. Three key findings emerge from our experiments: selecting prosody-encoding layers (6--12) from frozen Wav2Vec2 outperforms end-to-end finetuning (Score 0.207 vs. 0.161), as the non-verbal nature of BLEMORE audio makes phonetic layers irrelevant; the post-processing salience threshold $β$ varies from 0.05 to 0.43 across folds, revealing that personalized expression styles are the primary bottleneck; and task-adapted encoders collectively receive 62\% of ensemble weight over general-purpose baselines. Our 12-encoder system achieves Score = 0.279 (ACCP = 0.391, ACCS = 0.168) on the test set, placing 6th.

cs.CV↗

arXiv2026

TidyVoice 2026 Challenge Evaluation Plan

The performance of speaker verification systems degrades significantly under language mismatch, a critical challenge exacerbated by the field's reliance on English-centric data. To address this, we propose the TidyVoice Challenge for cross-lingual speaker verification. The challenge leverages the TidyVoiceX dataset from the novel TidyVoice benchmark, a large-scale, multilingual corpus derived from Mozilla Common Voice, and specifically curated to isolate the effect of language switching across approximately 40 languages. Participants will be tasked with building systems robust to this mismatch, with performance primarily evaluated using the Equal Error Rate on cross-language trials. By providing standardized data, open-source baselines, and a rigorous evaluation protocol, this challenge aims to drive research towards fairer, more inclusive, and language-independent speaker recognition technologies, directly aligning with the Interspeech 2026 theme, "Speaking Together."

eess.AS↗

arXiv2026

Adaptive Multimodal Person Recognition: A Robust Framework for Handling Missing Modalities

Person identification systems often rely on audio, visual, or behavioral cues, but real-world conditions frequently present with missing or degraded modalities. To address this challenge, we propose a multimodal person identification framework incorporating upper-body motion, face, and voice. Experimental results demonstrate that body motion outperforms traditional modalities such as face and voice in within-session evaluations, while serving as a complementary cue that enhances performance in multi-session scenarios. Our model employs a unified hybrid fusion strategy, fusing both feature-level and score-level information to maximize representational richness and decision accuracy. Specifically, it leverages multi-task learning to process modalities independently, followed by cross-attention and gated fusion mechanisms to exploit both unimodal information and cross-modal interactions. Finally, a confidence-weighted strategy and mistake-correction mechanism dynamically adapt to missing data, ensuring that our single classification head achieves optimal performance even in unimodal and bimodal scenarios. We evaluate our method on CANDOR, a newly introduced interview-based multimodal dataset, which we benchmark in this work for the first time. Our results demonstrate that the proposed trimodal system achieves 99.51% Top-1 accuracy on person identification tasks. In addition, we evaluate our model on the VoxCeleb1 dataset as a widely used evaluation protocol and reach 99.92% accuracy in bimodal mode, outperforming conventional approaches. Moreover, we show that our system maintains high accuracy even when one or two modalities are unavailable, making it a robust solution for real-world person recognition applications. The code and data for this work are publicly available.

cs.CV↗

arXiv2026

Spoofing-Aware Speaker Verification via Wavelet Prompt Tuning and Multi-Model Ensembles

This paper describes the UZH-CL system submitted to the SASV section of the WildSpoof 2026 challenge. The challenge focuses on the integrated defense against generative spoofing attacks by requiring the simultaneous verification of speaker identity and audio authenticity. We proposed a cascaded Spoofing-Aware Speaker Verification framework that integrates a Wavelet Prompt-Tuned XLSR-AASIST countermeasure with a multi-model ensemble. The ASV component utilizes the ResNet34, ResNet293, and WavLM-ECAPA-TDNN architectures, with Z-score normalization followed by score averaging. Trained on VoxCeleb2 and SpoofCeleb, the system obtained a Macro a-DCF of 0.2017 and a SASV EER of 2.08%. While the system achieved a 0.16% EER in spoof detection on the in-domain data, results on unseen datasets, such as the ASVspoof5, highlight the critical challenge of cross-domain generalization.

eess.AS↗

arXiv2026

TidyVoice: A Curated Multilingual Dataset for Speaker Verification Derived from Common Voice

The development of robust, multilingual speaker recognition systems is hindered by a lack of large-scale, publicly available and multilingual datasets, particularly for the read-speech style crucial for applications like anti-spoofing. To address this gap, we introduce the TidyVoice dataset derived from the Mozilla Common Voice corpus after mitigating its inherent speaker heterogeneity within the provided client IDs. TidyVoice currently contains training and test data from over 212,000 monolingual speakers (Tidy-M) and around 4,500 multilingual speakers (Tidy-X) from which we derive two distinct conditions. The Tidy-M condition contains target and non-target trials from monolingual speakers across 81 languages. The Tidy-X condition contains target and non-target trials from multilingual speakers in both same- and cross-language trials. We employ two architectures of ResNet models, achieving a 0.35% EER by fine-tuning on our comprehensive Tidy-M partition. Moreover, we show that this fine-tuning enhances the model's generalization, improving performance on unseen conversational interview data from the CANDOR corpus. The complete dataset, evaluation trials, and our models are publicly released to provide a new resource for the community.

eess.AS↗

arXiv2025

Towards Language-Independent Face-Voice Association with Multimodal Foundation Models

This paper describes the UZH-CL system submitted to the FAME2026 Challenge. The challenge focuses on cross-modal verification under unique multilingual conditions, specifically unseen and unheard languages. Our approach investigates two distinct architectures, consisting of a baseline dual-encoder system trained from scratch using contrastive and orthogonal projection losses, and a foundation model approach leveraging ImageBind with LoRA. To address the data scarcity and language constraints of the challenge, we curated an external Arabic dataset from VoxBlink. Our best-performing system, ImageBind-LoRA, demonstrates remarkable cross-lingual generalization: despite being fine-tuned exclusively on Arabic audio, it achieved an EER of 24.73% on the evaluation set (English and German), securing 2nd place in the competition.

eess.AS↗

arXiv2025

CL-UZH submission to the NIST SRE 2024 Speaker Recognition Evaluation

The CL-UZH team submitted one system each for the fixed and open conditions of the NIST SRE 2024 challenge. For the closed-set condition, results for the audio-only trials were achieved using the X-vector system developed with Kaldi. For the audio-visual results we used only models developed for the visual modality. Two sets of results were submitted for the open-set and closed-set conditions, one based on a pretrained model using the VoxBlink2 and VoxCeleb2 datasets. An Xvector-based model was trained from scratch using the CTS superset dataset for the closed set. In addition to the submission of the results of the SRE24 evaluation to the competition website, we talked about the performance of the proposed systems on the SRE24 evaluation in this report.

eess.AS↗

arXiv2025

State-Space Models in Efficient Whispered and Multi-dialect Speech Recognition

Whispered speech recognition presents significant challenges for conventional automatic speech recognition systems, particularly when combined with dialect variation. However, utilizing an efficient method to solve this problem using a low-range dataset and processing load is beneficial. This paper proposes a solution using a Mamba-based state-space model and four fine-tuned self-supervised models consisting of Wav2Vec2, WavLM, HuBERT, and Whisper to address the dual challenges of whispered speech and dialect diversity. Based on our knowledge, this represents the best performance reported on the wTIMIT and CHAINS datasets for whispered speech recognition. We trained the models using whispered and normal speech data across Singaporean, US, and Irish dialects. The findings demonstrated that utilizing the proposed Mamba-based model could work as a highly efficient model trained with low amounts of whispered data to simultaneously work on whispered and normal speech recognition. The code for this work is freely available.

eess.AS↗

arXiv2025

Quantifying and Reducing Speaker Heterogeneity within the Common Voice Corpus for Phonetic Analysis

With its crosslinguistic and cross-speaker diversity, the Mozilla Common Voice Corpus (CV) has been a valuable resource for multilingual speech technology and holds tremendous potential for research in crosslinguistic phonetics and speech sciences. Properly accounting for speaker variation is, however, key to the theoretical and statistical bases of speech research. While CV provides a client ID as an approximation to a speaker ID, multiple speakers can contribute under the same ID. This study aims to quantify and reduce heterogeneity in the client ID for a better approximation of a true, though still anonymous speaker ID. Using ResNet-based voice embeddings, we obtained a similarity score among recordings with the same client ID, then implemented a speaker discrimination task to identify an optimal threshold for reducing perceived speaker heterogeneity. These results have major downstream applications for phonetic analysis and the development of speaker-based speech technology.

eess.AS↗

arXiv2025

Multimodal Emotion Recognition and Sentiment Analysis in Multi-Party Conversation Contexts

Emotion recognition and sentiment analysis are pivotal tasks in speech and language processing, particularly in real-world scenarios involving multi-party, conversational data. This paper presents a multimodal approach to tackle these challenges on a well-known dataset. We propose a system that integrates four key modalities/channels using pre-trained models: RoBERTa for text, Wav2Vec2 for speech, a proposed FacialNet for facial expressions, and a CNN+Transformer architecture trained from scratch for video analysis. Feature embeddings from each modality are concatenated to form a multimodal vector, which is then used to predict emotion and sentiment labels. The multimodal system demonstrates superior performance compared to unimodal approaches, achieving an accuracy of 66.36% for emotion recognition and 72.15% for sentiment analysis.

cs.CV↗

arXiv2024

Leveraging Self-Supervised Models for 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% and a Character Error Rate (CER) of 4.24% on whispered speech. In contrast, the proposed WavLM-based system significantly improved performance, achieving a WER of 9.22% and a CER of 2.59%. 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.

eess.AS↗

arXiv2024

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.

eess.AS↗

arXiv2024

Gammatonegram Representation for End-to-End Dysarthric Speech Processing Tasks: Speech Recognition, Speaker Identification, and Intelligibility Assessment

Dysarthria is a disability that causes a disturbance in the human speech system and reduces the quality and intelligibility of a person's speech. Because of this effect, the normal speech processing systems can not work properly on impaired speech. This disability is usually associated with physical disabilities. Therefore, designing a system that can perform some tasks by receiving voice commands in the smart home can be a significant achievement. In this work, we introduce gammatonegram as an effective method to represent audio files with discriminative details, which is used as input for the convolutional neural network. On the other word, we convert each speech file into an image and propose image recognition system to classify speech in different scenarios. Proposed CNN is based on the transfer learning method on the pre-trained Alexnet. In this research, the efficiency of the proposed system for speech recognition, speaker identification, and intelligibility assessment is evaluated. According to the results on the UA dataset, the proposed speech recognition system achieved 91.29% accuracy in speaker-dependent mode, the speaker identification system acquired 87.74% accuracy in text-dependent mode, and the intelligibility assessment system achieved 96.47% accuracy in two-class mode. Finally, we propose a multi-network speech recognition system that works fully automatically. This system is located in a cascade arrangement with the two-class intelligibility assessment system, and the output of this system activates each one of the speech recognition networks. This architecture achieves an accuracy of 92.3% WRR. The source code of this paper is available.

eess.AS↗

arXiv2023

Facial Emotion Recognition Under Mask Coverage Using a Data Augmentation Technique

Identifying human emotions using AI-based computer vision systems, when individuals wear face masks, presents a new challenge in the current Covid-19 pandemic. In this study, we propose a facial emotion recognition system capable of recognizing emotions from individuals wearing different face masks. A novel data augmentation technique was utilized to improve the performance of our model using four mask types for each face image. We evaluated the effectiveness of four convolutional neural networks, Alexnet, Squeezenet, Resnet50 and VGGFace2 that were trained using transfer learning. The experimental findings revealed that our model works effectively in multi-mask mode compared to single-mask mode. The VGGFace2 network achieved the highest accuracy rate, with 97.82% for the person-dependent mode and 74.21% for the person-independent mode using the JAFFE dataset. However, we evaluated our proposed model using the UIBVFED dataset. The Resnet50 has demonstrated superior performance, with accuracies of 73.68% for the person-dependent mode and 59.57% for the person-independent mode. Moreover, we employed metrics such as precision, sensitivity, specificity, AUC, F1 score, and confusion matrix to measure our system's efficiency in detail. Additionally, the LIME algorithm was used to visualize CNN's decision-making strategy.

cs.CV↗