An official website of the United States government
When dealing with overlapped speech, the performance of automatic speech recognition (ASR) systems substantially degrades as they are designed for single-talker speech. To enhance ASR performance in conversational or meeting environments, continuous speaker separation (CSS) is commonly employed. However, CSS requires a short separation window to avoid many speakers inside the window and sequential grouping of discontinuous speech segments. To address these limitations, we introduce a new multi-channel framework called “speaker separation via neural diarization” (SSND) for meeting environments. Our approach utilizes an end-to-end diarization system to identify the speech activity of each individual speaker. By leveraging estimated speaker boundaries, we generate a sequence of embeddings, which in turn facilitate the assignment of speakers to the outputs of a multi-talker separation model. SSND addresses the permutation ambiguity issue of talker-independent speaker separation during the diarization phase through location-based training, rather than during the separation process. This unique approach allows multiple non-overlapped speakers to be assigned to the same output stream, making it possible to efficiently process long segments—a task impossible with CSS. Additionally, SSND is naturally suitable for speaker-attributed ASR. We evaluate our proposed diarization and separation methods on the open LibriCSS dataset, advancing state-of-the-art diarization and ASR results by a large margin.
more »
« less
This content will become publicly available on July 16, 2026
A brain-inspired algorithm enhances automatic speech recognition performance in multi-talker scenes
Abstract Modern automatic speech recognition (ASR) systems are capable of impressive performance recognizing clean speech but struggle in noisy, multi-talker environments, commonly referred to as the “cocktail party problem.” In contrast, many human listeners can solve this problem, suggesting the existence of a solution in the brain. Here we present a novel approach that uses a brain inspired sound segregation algorithm (BOSSA) as a preprocessing step for a state-of-the-art ASR system (Whisper). We evaluated BOSSA’s impact on ASR accuracy in a spatialized multi-talker scene with one target speaker and two competing maskers, varying the difficulty of the task by changing the target-to-masker ratio. We found that median word error rate improved by up to 54% when the target-to-masker ratio was low. Our results indicate that brain-inspired algorithms have the potential to considerably enhance ASR accuracy in challenging multi-talker scenarios without the need for retraining or fine-tuning existing state-of-the-art ASR systems.
more »
« less
- Award ID(s):
- 2319321
- PAR ID:
- 10617065
- Publisher / Repository:
- bioRxiv
- Date Published:
- Format(s):
- Medium: X
- Institution:
- bioRxiv
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
The performance of automatic speech recognition (ASR) systems severely degrades when multi-talker speech overlap occurs. In meeting environments, speech separation is typically performed to improve the robustness of ASR systems. Recently, location-based training (LBT) was proposed as a new training criterion for multi-channel talker-independent speaker separation. Assuming fixed array geometry, LBT outperforms widely-used permutation-invariant training in fully overlapped utterances and matched reverberant conditions. This paper extends LBT to conversational multi-channel speaker separation. We introduce multi-resolution LBT to estimate the complex spectrograms from low to high time and frequency resolutions. With multi-resolution LBT, convolutional kernels are assigned consistently based on speaker locations in physical space. Evaluation results show that multi-resolution LBT consistently outperforms other competitive methods on the recorded LibriCSS corpus.more » « less
-
Measuring how well human listeners recognize speech under varying environmental conditions (speech intelligibility) is a challenge for theoretical, technological, and clinical approaches to speech communication. The current gold standard—human transcription—is time- and resource-intensive. Recent advances in automatic speech recognition (ASR) systems raise the possibility of automating intelligibility measurement. This study tested 4 state-of-the-art ASR systems with second language speech-in-noise and found that one, whisper, performed at or above human listener accuracy. However, the content of whisper's responses diverged substantially from human responses, especially at lower signal-to-noise ratios, suggesting both opportunities and limitations for ASR--based speech intelligibility modeling.more » « less
-
Abstract Multilingual speakers can find speech recognition in everyday environments like restaurants and open-plan offices particularly challenging. In a world where speaking multiple languages is increasingly common, effective clinical and educational interventions will require a better understanding of how factors like multilingual contexts and listeners’ language proficiency interact with adverse listening environments. For example, word and phrase recognition is facilitated when competing voices speak different languages. Is this due to a “release from masking” from lower-level acoustic differences between languages and talkers, or higher-level cognitive and linguistic factors? To address this question, we created a “one-man bilingual cocktail party” selective attention task using English and Mandarin speech from one bilingual talker to reduce low-level acoustic cues. In Experiment 1, 58 listeners more accurately recognized English targets when distracting speech was Mandarin compared to English. Bilingual Mandarin–English listeners experienced significantly more interference and intrusions from the Mandarin distractor than did English listeners, exacerbated by challenging target-to-masker ratios. In Experiment 2, 29 Mandarin–English bilingual listeners exhibited linguistic release from masking in both languages. Bilinguals experienced greater release from masking when attending to English, confirming an influence of linguistic knowledge on the “cocktail party” paradigm that is separate from primarily energetic masking effects. Effects of higher-order language processing and expertise emerge only in the most demanding target-to-masker contexts. The “one-man bilingual cocktail party” establishes a useful tool for future investigations and characterization of communication challenges in the large and growing worldwide community of Mandarin–English bilinguals.more » « less
-
Speech is a natural channel for human-computer interaction in robotics and consumer applications. Natural language understanding pipelines that start with speech can have trouble recovering from speech recognition errors. Black-box automatic speech recognition (ASR) systems, built for general purpose use, are unable to take advantage of in-domain language models that could otherwise ameliorate these errors. In this work, we present a method for re-ranking black-box ASR hypotheses using an in-domain language model and semantic parser trained for a particular task. Our re-ranking method significantly improves both transcription accuracy and semantic understanding over a state-of-the-art ASR’s vanilla output.more » « less
