Selective Attention System (SAS): Device-Addressed Speech Detection for Real-Time On-Device Voice AI

• URL: http://arxiv.org/abs/2604.08412v1
• Date: Thu, 09 Apr 2026 16:11:15 GMT
• Title: Selective Attention System (SAS): Device-Addressed Speech Detection for Real-Time On-Device Voice AI
• Abstract summary: We show that, in multi-speaker environments with temporally ambiguous utterances, this task is more effectively modelled as a sequential routing problem over interaction history than as an utterance-local classification task.<n>We formalize this as Sequential Device-Addressed Routing (SDAR) and present the Selective Attention System (SAS), an on-device implementation that instantiates this formulation.
• Score: 1.399948157377307
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: We study device-addressed speech detection under pre-ASR edge deployment constraints, where systems must decide whether to forward audio before transcription under strict latency and compute limits. We show that, in multi-speaker environments with temporally ambiguous utterances, this task is more effectively modelled as a sequential routing problem over interaction history than as an utterance-local classification task. We formalize this as Sequential Device-Addressed Routing (SDAR) and present the Selective Attention System (SAS), an on-device implementation that instantiates this formulation. On a held-out 60-hour multi-speaker English test set, the primary audio-only configuration achieves F1=0.86 (precision=0.89, recall=0.83); with an optional camera, audio+video fusion raises F1 to 0.95 (precision=0.97, recall=0.93). Removing causal interaction history (Stage~3) reduced F1 from 0.95 to 0.57+/-0.03 in the audio+video configuration under our evaluation protocol. Among the tested components, this was the largest observed ablation effect, indicating that short-horizon interaction history carries substantial decision-relevant information in the evaluated setting. SAS runs fully on-device on ARM Cortex-A class hardware (<150 ms latency, <20 MB footprint). All results are from internal evaluation on a proprietary dataset evaluated primarily in English; a 5-hour evaluation subset may be shared for independent verification (Section 8.8).

Related papers

• A Hierarchical End-of-Turn Model with Primary Speaker Segmentation for Real-Time Conversational AI [3.995809098639418]
  We present a real-time front-end for voice-based conversational AI to enable natural turn-taking in two-speaker scenarios.<n>To operate robustly in multi-speaker environments, the system continuously identifies and tracks the primary user.
  arXiv  Detail & Related papers  (2026-03-10T21:17:17Z)
• Stream RAG: Instant and Accurate Spoken Dialogue Systems with Streaming Tool Usage [66.67531241554546]
  End-to-end speech-in speech-out dialogue systems are emerging as a powerful alternative to traditional ASR-LLM-TTS pipelines.<n>We introduce the first approach to extend tool use directly into speech-in speech-out systems.<n>We propose Streaming Retrieval-Augmented Generation (Streaming RAG), a novel framework that reduces user-perceived latency by predicting tool queries in parallel with user speech.
  arXiv  Detail & Related papers  (2025-10-02T14:18:20Z)
• Detect Any Sound: Open-Vocabulary Sound Event Detection with Multi-Modal Queries [18.147981850263708]
  We propose a query-based framework for open-vocabulary SED guided by multi-modal queries.<n>DASM formulates SED as a frame-level retrieval task, where audio features are matched against query vectors from text or audio prompts.<n>DASM effectively balances localization accuracy with generalization to novel classes, outperforming CLAP-based methods in open-vocabulary setting.
  arXiv  Detail & Related papers  (2025-07-22T08:24:01Z)
• End-to-End Integration of Speech Separation and Voice Activity Detection for Low-Latency Diarization of Telephone Conversations [13.020158123538138]
  Speech separation guided diarization (SSGD) performs diarization by first separating the speakers and then applying voice activity detection (VAD) on each separated stream. We consider three state-of-the-art speech separation (SSep) algorithms and study their performance in online and offline scenarios. We show that our best model achieves 8.8% DER on CALLHOME, which outperforms the current state-of-the-art end-to-end neural diarization model.
  arXiv  Detail & Related papers  (2023-03-21T16:33:56Z)
• Continual Learning for On-Device Speech Recognition using Disentangled Conformers [54.32320258055716]
  We introduce a continual learning benchmark for speaker-specific domain adaptation derived from LibriVox audiobooks. We propose a novel compute-efficient continual learning algorithm called DisentangledCL. Our experiments show that the DisConformer models significantly outperform baselines on general ASR.
  arXiv  Detail & Related papers  (2022-12-02T18:58:51Z)
• Unified End-to-End Speech Recognition and Endpointing for Fast and Efficient Speech Systems [17.160006765475988]
  We propose a method to jointly train the ASR and EP tasks in a single end-to-end (E2E) model. We introduce a "switch" connection, which trains the EP to consume either the audio frames directly or low-level latent representations from the ASR model. This results in a single E2E model that can be used during inference to perform frame filtering at low cost.
  arXiv  Detail & Related papers  (2022-11-01T23:43:15Z)
• Rethinking Audio-visual Synchronization for Active Speaker Detection [62.95962896690992]
  Existing research on active speaker detection (ASD) does not agree on the definition of active speakers. We propose a cross-modal contrastive learning strategy and apply positional encoding in attention modules for supervised ASD models to leverage the synchronization cue. Experimental results suggest that our model can successfully detect unsynchronized speaking as not speaking, addressing the limitation of current models.
  arXiv  Detail & Related papers  (2022-06-21T14:19:06Z)
• End-to-End Active Speaker Detection [58.7097258722291]
  We propose an end-to-end training network where feature learning and contextual predictions are jointly learned. We also introduce intertemporal graph neural network (iGNN) blocks, which split the message passing according to the main sources of context in the ASD problem. Experiments show that the aggregated features from the iGNN blocks are more suitable for ASD, resulting in state-of-the art performance.
  arXiv  Detail & Related papers  (2022-03-27T08:55:28Z)
• HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units [81.53783563025084]
  We propose an offline clustering step to provide aligned target labels for a BERT-like prediction loss. A key ingredient of our approach is applying the prediction loss over the masked regions only. HuBERT shows up to 19% and 13% relative WER reduction on the more challenging dev-other and test-other evaluation subsets.
  arXiv  Detail & Related papers  (2021-06-14T14:14:28Z)
• Dissecting User-Perceived Latency of On-Device E2E Speech Recognition [34.645194215436966]
  We show that factors affecting token emission latency, and endpointing behavior significantly impact on user-perceived latency (UPL) We achieve the best trade-off between latency and word error rate when performing ASR jointly with endpointing, and using the recently proposed alignment regularization.
  arXiv  Detail & Related papers  (2021-04-06T00:55:11Z)
• End-to-End Automatic Speech Recognition Integrated With CTC-Based Voice Activity Detection [48.80449801938696]
  This paper integrates a voice activity detection function with end-to-end automatic speech recognition. We focus on connectionist temporal classification ( CTC) and its extension ofsynchronous/attention. We use the labels as a cue for detecting speech segments with simple thresholding.
  arXiv  Detail & Related papers  (2020-02-03T03:36:34Z)
• Continuous speech separation: dataset and analysis [52.10378896407332]
  In natural conversations, a speech signal is continuous, containing both overlapped and overlap-free components. This paper describes a dataset and protocols for evaluating continuous speech separation algorithms.
  arXiv  Detail & Related papers  (2020-01-30T18:01:31Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.