Audio-visual multimodal speech recognition method fusing audio and video data

Audio-visual speech recognition (AVSR) is one of the most promising solutions for reliable speech recognition, particularly when audio is corrupted by noise. Additional visual information can be used for both automatic lip-reading and gesture recognition. Hand gestures are a form of non-verbal communication and can be used as a very important part of modern human–computer interaction systems. Currently, audio and video modalities are easily accessible by sensors of mobile devices. However, there is no out-of-the-box solution for automatic audio-visual speech and gesture recognition. This study introduces two deep neural network-based model architectures: one for AVSR and one for gesture recognition. The main novelty regarding audio-visual speech recognition lies in fine-tuning strategies for both visual and acoustic features and in the proposed end-to-end model, which considers three modality fusion approaches: prediction-level, feature-level, and model-level. The main novelty in gesture recognition lies in a unique set of spatio-temporal features, including those that consider lip articulation information. As there are no available datasets for the combined task, we evaluated our methods on two different large-scale corpora—LRW and AUTSL—and outperformed existing methods on both audio-visual speech recognition and gesture recognition tasks. We achieved AVSR accuracy for the LRW dataset equal to 98.76% and gesture recognition rate for the AUTSL dataset equal to 98.56%. The results obtained demonstrate not only the high performance of the proposed methodology, but also the fundamental possibility of recognizing audio-visual speech and gestures by sensors of mobile devices.

We describe an audio-visual automatic continuous speech recognition system, which significantly improves speech recognition performance over a wide range of acoustic noise levels, as well as under clean audio conditions. The system utilizes facial animation parameters (FAPs) supported by the MPEG-4 standard for the visual representation of speech. We also describe a robust and automatic algorithm we have developed to extract FAPs from visual data, which does not require hand labeling or extensive training procedures. The principal component analysis (PCA) was performed on the FAPs in order to decrease the dimensionality of the visual feature vectors, and the derived projection weights were used as visual features in the audio-visual automatic speech recognition (ASR) experiments. Both single-stream and multistream hidden Markov models (HMMs) were used to model the ASR system, integrate audio and visual information, and perform a relatively large vocabulary (approximately 1000 words) speech recognition experiments. The experiments performed use clean audio data and audio data corrupted by stationary white Gaussian noise at various SNRs. The proposed system reduces the word error rate (WER) by 20% to 23% relatively to audio-only speech recognition WERs, at various SNRs (0–30 dB) with additive white Gaussian noise, and by 19% relatively to audio-only speech recognition WER under clean audio conditions.

Automatic speech recognition (ASR) has gained wide popularity in last decade. Various devices like mobile phones, computers, vehicles, and audio/video players are now being equipped with ASR technology. The increasing use and dependence on ASR technology leads to research enhancements and opportunities in this domain. This chapter provides a detailed review of various advancements in ASR systems development. It highlights history of speech recognition followed by detailed insight into recent advancements and industry leaders providing latest solutions. ASR framework has been discussed in detail which includes feature extraction techniques, acoustic modeling techniques, and language modeling techniques. The chapter also lists various popular data sets available and discusses generation of new data sets. This work will be helpful for the researchers who are new to this field and are exploring development of new speech recognition techniques.

Combined speech enhancement and auditory modelling for robust distributed speech recognition

Human computer interaction (HCI) is very crucial in our day-to-day activity. Speech is one of the essential and intuitive ways to interact with machines such as Smartphone, which has multiple sensors as microphone, camera, etc. An efficient performance speech recognition system improves interaction between man and machines by making latter more receptive to user needs. Such system has Automatic speech recognition (ASR) engine, which is facing a unique challenge of accuracy in recognition rate. By integrating acoustic signal feature vectors with the visual features, a more robust audiovisual speech recognition engine (AVSR) could be developed for real environmental scenarios. This paper presents past research and development in the field of ASR and AVSR technologies. It describes key technological perspective and admiration of the fundamental progress in ASR and AVSR. The objective of this review is to summarize and compare some of the well-known methods experimented by previous researchers, and to conclude with direction on future research proficiency in HCI system using ASR and AVSR engine.

To date, multimodal speech recognition systems based on the processing of audio and video signals show significantly better results than their unimodal counterparts. In general, researchers divide the solution of the audio–visual speech recognition problem into two parts. First, in extracting the most informative features from each modality and second, in the most successful way of fusion both modalities. Ultimately, this leads to an improvement in the accuracy of speech recognition. Almost all modern studies use this approach with video data of a standard recording speed of 25 frames per second. The choice of such a recording speed is easily explained, since the vast majority of existing audio–visual databases are recorded with this rate. However, it should be noticed that the number of 25 frames per second is a world standard for many areas and has never been specifically calculated for speech recognition tasks. The main purpose of this study is to investigate the effect brought by the high-speed video data (up to 200 frames per second) on the speech recognition accuracy. And also to find out whether the use of a high-speed video camera makes the speech recognition systems more robust to acoustical noise. To this end, we recorded a database of audio–visual Russian speech with high-speed video recordings, which consists of records of 20 speakers, each of them pronouncing 200 phrases of continuous Russian speech. Experiments performed on this database showed an improvement in the absolute speech recognition rate up to 3.10%. We also proved that the use of the high-speed camera with 200 fps allows achieving better recognition results under different acoustically noisy conditions (signal-to-noise ratio varied between 40 and 0 dB) with different types of noise (e.g. white noise, babble noise).

Purpose: High-speed videoendoscopy (HSV) is an emerging, but barely used, endoscopy technique in the clinic to assess and diagnose voice disorders because of the lack of dedicated software to analyze the data. HSV allows to quantify the vocal fold oscillations by segmenting the glottal area. This challenging task has been tackled by various studies; however, the proposed approaches are mostly limited and not suitable for daily clinical routine.Method: We developed a user-friendly software in C# that allows the editing, motion correction, segmentation, and quantitative analysis of HSV data. We further provide pretrained deep neural networks for fully automatic glottis segmentation.Results: We freely provide our software Glottis Analysis Tools (GAT). Using GAT, we provide a general threshold-based region growing platform that enables the user to analyze data from various sources, such as in vivo recordings, ex vivo recordings, and high-speed footage of artificial vocal folds. Additionally, especially for in vivo recordings, we provide three robust neural networks at various speed and quality settings to allow a fully automatic glottis segmentation needed for application by untrained personnel. GAT further evaluates video and audio data in parallel and is able to extract various features from the video data, among others the glottal area waveform, that is, the changing glottal area over time. In total, GAT provides 79 unique quantitative analysis parameters for video- and audio-based signals. Many of these parameters have already been shown to reflect voice disorders, highlighting the clinical importance and usefulness of the GAT software.Conclusion: GAT is a unique tool to process HSV and audio data to determine quantitative, clinically relevant parameters for research, diagnosis, and treatment of laryngeal disorders.Supplemental Material S1. Supplemental figures: Figure S1. Neural network training metrics.Figure S2. Neural network evaluation.Figure S3. Example segmentations using different backbones from various source data.Figure S4. Audio analysis platform.Supplemental Material S2. Endoscopic footage with semi-automatic and fully automatic segmentation methods, multiple examples.Supplemental Material S3. GAT manual. Kist, A. M., Gomez, P., Dubrovskiy, D., Schlegel, P., Kunduk, M., Echternach, M., Patel, R., Semmler, M., Bohr, C., Durr, S., Schutzenberger, A., & Dollinger, M. (2021). A deep learning enhanced novel software tool for laryngeal dynamics analysis. Journal of Speech, Language, and Hearing Research. Advance online publication. https://doi.org/10.1044/2021_JSLHR-20-00498

Abstract— Audio data carries rich information in the form of speech, music, and environmental sounds, but its raw waveform is often complex and high-dimensional, making direct analysis difficult. Neural network architectures have emerged as powerful tools for extracting meaningful representations from audio signals, enabling efficient analysis and interpretation. Recurrent Neural Networks (RNNs), and Transformer-based models— for learning robust and discriminative features from audio. By automatically capturing temporal, spectral, and contextual patterns, these architectures significantly improve performance in tasks such as speech recognition, speaker identification, music classification, and environmental sound detection. The findings highlight the potential of neural networks to replace traditional handcrafted features, thereby advancing the development of scalable, accurate, and realtime audio processing applications. The rapid growth of audio data across domains such as speech, music, healthcare, and environmental monitoring has created a strong need for effective methods to extract meaningful representations from complex audio signals. Traditional approaches rely on handcrafted features like MFCCs and spectrogram descriptors, which often fail to capture the full temporal and spectral dynamics present in raw audio. Keywords: Neural Networks, Deep Learning, Audio Representation Learning, Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Transformer Models, Feature Extraction, Speech Recognition, Speaker Identification, Sound Event Detection, Spectrogram Analysis, Audio Signal Processing

The impact of automatic exaggeration of the visual articulatory features of a talker on the intelligibility of spectrally distorted speech

HMM-based audio-visual speech recognition (AVSR) systems have shown success in continuous speech recognition by combining visual and audio information, especially in noisy environments. In this paper we study how to improve decision trees used to create context classes in HMM-based AVSR systems. Traditionally, visual models have been trained with the same context classes as the audio only models. In this paper we investigate the use of separate decision trees to model the context classes for the audio and visual streams independently. Additionally we investigate the use of viseme classes in the decision tree building for the visual stream. On experiments with a 37-speaker 1.5 hours test set (about 12000 words) of continuous digits in noise, we obtain about a 3% absolute (20% relative) gain on AVSR performance by using separate decision trees for the audio and visual streams when using viseme classes in decision tree building for the visual stream.

In this paper, we present a new technique for social mobile robot speech recognition based on scene-associated training models. The key contribution of the paper is a real-time framework that reduces the effect of room reverberation and ambient noise, a challenging problem in speech recognition. In classical approaches, anechoic sound is used to train the model, with the main focus on removing reverberation or noise from the sound. Our technique differs in that we train a number of speech recognizers directly from the reverberated sound, by associating each recognizer with a unique visual scene, to deal with the varying reverberation properties of different rooms. By extracting local features from a captured image and recognizing a scene, the robot can use the appropriate speech recognizer that is trained for the particular structural properties of that scene. We tested our method by using a baseline speech recognition model (HTK) across a variety of rooms and different levels of background noise. The results show that the association between a visual scene and a corresponding speech recognizer greatly improves the robot's speech recognition accuracy, together with increasing the computational speed of recognition, compared to competing techniques.

There has been a long term interest in using visual information to improve automatic speech recognition (ASR) system performance. Both audio and visual information are required in conventional audio visual speech recognition (AVSR) systems. This limits their wider applications when visual modality is not present. To this end, one possible solution is to use acoustic-to-visual (A2V) inversion techniques to generate visual features. Previous research in this direction used synthetic acoustic-articulatory parallel data in inversion model training. The acoustic mismatch between the audio-visual (AV) parallel data and target data was not considered. In addition, the target language to apply these technologies has been focused on English. In this article, a real 3D Audio-Visual Mandarin Continuous Speech (3DAV-MCS) corpus was used to train deep neural network based A2V inversion models. Cross-domain adaptation of the inversion models allows suitable visual features to be generated from acoustic data of mismatched domains. The proposed cross-domain deep visual feature generation techniques were evaluated on two state-of-the-art Mandarin speech recognition tasks: DAPRA GALE broadcast transcription and BOLT conversational telephone speech recognition. The AVSR systems constructed using the cross-domain generated visual features consistently outperformed the baseline convolutional neural network (CNN) ASR systems by up to 3.3% absolute (9.1% relative) character error rate (CER) reductions after both speaker adaptive training and sequence discriminative training were performed.

Audio-visual speech recognition is a natural and robust approach to improving human-robot interaction in noisy environments. Although multi-stream Dynamic Bayesian Network and coupled HMM are widely used for audio-visual speech recognition, they fail to learn the shared features between modalities and ignore the dependency of features among the frames within each discrete state. In this paper, we propose a Deep Dynamic Bayesian Network (DDBN) to perform unsupervised extraction of spatial-temporal multimodal features from Tibetan audio-visual speech data and build an accurate audio-visual speech recognition model under a no frame-independency assumption. The experiment results on Tibetan speech data from some real-world environments showed the proposed DDBN outperforms the state-of-art methods in word recognition accuracy.

This study is part of an ongoing collaborative effort between the medical and the signal processing communities to promote research on applying standard Automatic Speech Recognition (ASR) techniques for the automatic diagnosis of patients with severe obstructive sleep apnoea (OSA). Early detection of severe apnoea cases is important so that patients can receive early treatment. Effective ASR-based detection could dramatically cut medical testing time. Working with a carefully designed speech database of healthy and apnoea subjects, we describe an acoustic search for distinctive apnoea voice characteristics. We also study abnormal nasalization in OSA patients by modelling vowels in nasal and nonnasal phonetic contexts using Gaussian Mixture Model (GMM) pattern recognition on speech spectra. Finally, we present experimental findings regarding the discriminative power of GMMs applied to severe apnoea detection. We have achieved an 81% correct classification rate, which is very promising and underpins the interest in this line of inquiry.

Speech recognition, the transformation of spoken language into written text, is becoming increasingly vital across a broad range of applications. Despite the advancements in end-to-end Neural Network (NN) based speech recognition systems, the requirement for large volumes of annotated audio data tailored to specific scenarios remains a significant challenge. To address this, we introduce a novel approach, the Character Audio Mix-up (CAMP), which synthesizes scenario-specific audio data for Mandarin at a significantly reduced cost and effort. This method concatenates the audio segments of each character’s Pinyin in the text, obtained through force alignment on an existing annotated dataset, to synthesize the audio. These synthesized audios are then used to train the Automatic Speech Recognition (ASR) models. Experiments conducted on the AISHELL-3, and AIDATATANG datasets validate the effectiveness of CAMP, with ASR models trained on CAMP synthesized data performing relatively well compared to those trained with actual data from these datasets. Further, our ablation study reveals that while synthesized audio data can significantly reduce the need for real annotated audio specific to each scenario, it cannot entirely replace real audio. Thus, the importance of real annotated audio data in specific application scenarios is emphasized.