Abstract:This paper presents a comprehensive chatbot system designed to handle a wide range of audio-related queries by integrating multiple specialized audio processing models. The proposed system uses an intent classifier, trained on a diverse audio query dataset, to route queries about audio content to expert models such as Automatic Speech Recognition (ASR), Speaker Diarization, Music Identification, and Text-to-Audio generation. A 3.8 B LLM model then takes inputs from an Audio Context Detection (ACD) module extracting audio event information from the audio and post processes text domain outputs from the expert models to compute the final response to the user. We evaluated the system on custom audio tasks and MMAU sound set benchmarks. The custom datasets were motivated by target use cases not covered in industry benchmarks and included ACD-timestamp-QA (Question Answering) as well as ACD-temporal-QA datasets to evaluate timestamp and temporal reasoning questions, respectively. First we determined that a BERT based Intent Classifier outperforms LLM-fewshot intent classifier in routing queries. Experiments further show that our approach significantly improves accuracy on some custom tasks compared to state-of-the-art Large Audio Language Models and outperforms models in the 7B parameter size range on the sound testset of the MMAU benchmark, thereby offering an attractive option for on device deployment.
Abstract:Systems that automatically generate text captions for audio, images and video lack a confidence indicator of the relevance and correctness of the generated sequences. To address this, we build on existing methods of confidence measurement for text by introduce selective pooling of token probabilities, which aligns better with traditional correctness measures than conventional pooling does. Further, we propose directly measuring the similarity between input audio and text in a shared embedding space. To measure self-consistency, we adapt semantic entropy for audio captioning, and find that these two methods align even better than pooling-based metrics with the correctness measure that calculates acoustic similarity between captions. Finally, we explain why temperature scaling of confidences improves calibration.
Abstract:Noise suppression (NS) algorithms are effective in improving speech quality in many cases. However, aggressive noise suppression can damage the target speech, reducing both speech intelligibility and quality despite removing the noise. This study proposes an explicit speech restoration method using a voice conversion (VC) technique for restoration after noise suppression. We observed that high-quality speech can be restored through a diffusion-based voice conversion stage, conditioned on the target speaker embedding and speech content information extracted from the de-noised speech. This speech restoration can achieve enhancement effects such as bandwidth extension, de-reverberation, and in-painting. Our experimental results demonstrate that this two-stage NS+VC framework outperforms single-stage enhancement models in terms of output speech quality, as measured by objective metrics, while scoring slightly lower in speech intelligibility. To further improve the intelligibility of the combined system, we propose a content encoder adaptation method for robust content extraction in noisy conditions.
Abstract:The Audio Question Answering task includes audio event classification, audio captioning, and open ended reasoning. Recently, Audio Question Answering has garnered attention due to the advent of Large Audio Language Models. Current literature focuses on constructing LALMs by integrating audio encoders with text only Large Language Models through a projection module. While Large Audio Language Models excel in general audio understanding, they are limited in temporal reasoning which may hinder their commercial applications and on device deployment. This paper addresses these challenges and limitations in audio temporal reasoning. First, we introduce a data augmentation technique for generating reliable audio temporal questions and answers using an LLM. Second, we propose a continued finetuning curriculum learning strategy to specialize in temporal reasoning without compromising performance on finetuned tasks. Finally, we develop a reliable and transparent automated metric, assisted by an LLM, to measure the correlation between Large Audio Language Model responses and ground truth data intelligently. We demonstrate the effectiveness of our proposed techniques using SOTA LALMs on public audio benchmark datasets.
Abstract:While many recent any-to-any voice conversion models succeed in transferring some target speech's style information to the converted speech, they still lack the ability to faithfully reproduce the speaking style of the target speaker. In this work, we propose a novel method to extract rich style information from target utterances and to efficiently transfer it to source speech content without requiring text transcriptions or speaker labeling. Our proposed approach introduces an attention mechanism utilizing a self-supervised learning (SSL) model to collect the speaking styles of a target speaker each corresponding to the different phonetic content. The styles are represented with a set of embeddings called stylebook. In the next step, the stylebook is attended with the source speech's phonetic content to determine the final target style for each source content. Finally, content information extracted from the source speech and content-dependent target style embeddings are fed into a diffusion-based decoder to generate the converted speech mel-spectrogram. Experiment results show that our proposed method combined with a diffusion-based generative model can achieve better speaker similarity in any-to-any voice conversion tasks when compared to baseline models, while the increase in computational complexity with longer utterances is suppressed.
Abstract:We propose a highly controllable voice manipulation system that can perform any-to-any voice conversion (VC) and prosody modulation simultaneously. State-of-the-art VC systems can transfer sentence-level characteristics such as speaker, emotion, and speaking style. However, manipulating the frame-level prosody, such as pitch, energy and speaking rate, still remains challenging. Our proposed model utilizes a frame-level prosody feature to effectively transfer such properties. Specifically, pitch and energy trajectories are integrated in a prosody conditioning module and then fed alongside speaker and contents embeddings to a diffusion-based decoder generating a converted speech mel-spectrogram. To adjust the speaking rate, our system includes a self-supervised model based post-processing step which allows improved controllability. The proposed model showed comparable speech quality and improved intelligibility compared to a SOTA approach. It can cover a varying range of fundamental frequency (F0), energy and speed modulation while maintaining converted speech quality.
Abstract:There has been significant research on developing pretrained transformer architectures for multimodal-to-text generation tasks. Albeit performance improvements, such models are frequently overparameterized, hence suffer from hallucination and large memory footprint making them challenging to deploy on edge devices. In this paper, we address both these issues for the application of automated audio captioning. First, we propose a data augmentation technique for generating hallucinated audio captions and show that similarity based on an audio-text shared latent space is suitable for detecting hallucination. Then, we propose a parameter efficient inference time faithful decoding algorithm that enables smaller audio captioning models with performance equivalent to larger models trained with more data. During the beam decoding step, the smaller model utilizes an audio-text shared latent representation to semantically align the generated text with corresponding input audio. Faithful guidance is introduced into the beam probability by incorporating the cosine similarity between latent representation projections of greedy rolled out intermediate beams and audio clip. We show the efficacy of our algorithm on benchmark datasets and evaluate the proposed scheme against baselines using conventional audio captioning and semantic similarity metrics while illustrating tradeoffs between performance and complexity.
Abstract:Advancement in large pretrained language models has significantly improved their performance for conditional language generation tasks including summarization albeit with hallucinations. To reduce hallucinations, conventional methods proposed improving beam search or using a fact checker as a postprocessing step. In this paper, we investigate the use of the Natural Language Inference (NLI) entailment metric to detect and prevent hallucinations in summary generation. We propose an NLI-assisted beam re-ranking mechanism by computing entailment probability scores between the input context and summarization model-generated beams during saliency-enhanced greedy decoding. Moreover, a diversity metric is introduced to compare its effectiveness against vanilla beam search. Our proposed algorithm significantly outperforms vanilla beam decoding on XSum and CNN/DM datasets.
Abstract:Model architectures such as wav2vec 2.0 and HuBERT have been proposed to learn speech representations from audio waveforms in a self-supervised manner. When these models are combined with downstream tasks such as speech recognition, they have been shown to provide state-of-the-art performance. However, these models use a large number of parameters, the smallest version of which has about 95 million parameters. This constitutes a challenge for edge AI device deployments. In this paper, we use knowledge distillation to reduce the original model size by about 75% while maintaining similar performance levels. Moreover, we use wav2vec 2.0 and HuBERT models for distillation and present a comprehensive performance analysis through our experiments where we fine-tune the distilled models on single task and multi-task frameworks separately. In particular, our experiments show that fine-tuning the distilled models on keyword spotting and speaker verification tasks result in only 0.1% accuracy and 0.9% equal error rate degradations, respectively.
Abstract:In the era of loT (Internet of Things) we are surrounded by a plethora of Al enabled devices that can transcribe images, video, audio, and sensors signals into text descriptions. When such transcriptions are captured in activity reports for monitoring, life logging and anomaly detection applications, a user would typically request a summary or ask targeted questions about certain sections of the report they are interested in. Depending on the context and the type of question asked, a question answering (QA) system would need to automatically determine whether the answer covers single-span or multi-span text components. Currently available QA datasets primarily focus on single span responses only (such as SQuAD[4]) or contain a low proportion of examples with multiple span answers (such as DROP[3]). To investigate automatic selection of single/multi-span answers in the use case described, we created a new smart home environment dataset comprised of questions paired with single-span or multi-span answers depending on the question and context queried. In addition, we propose a RoBERTa[6]-based multiple span extraction question answering (MSEQA) model returning the appropriate answer span for a given question. Our experiments show that the proposed model outperforms state-of-the-art QA models on our dataset while providing comparable performance on published individual single/multi-span task datasets.