Spectron is designed to perform question answering (QA) and speech continuation using a pre-trained speech encoder and a large language model (LLM). The model operates end-to-end on spectrograms, ...

A novel approach adapts pre-trained LLMs for question answering and speech continuation by incorporating a pre-trained speech encoder. This enables the model to handle speech inputs and outputs. The ...

Abstract: Acoustic features play an important role in improving the quality of the synthesised speech. Currently, the Mel spectrogram is a widely employed acoustic feature in most acoustic models.

Abstract: In this study, we extend the capability of the method of relative-to-maximum masking (RMM) in speech enhancement by further leveraging the importance of each time-frequency unit in the ...

Speech loss due to neurological deficits is a severe disability that limits both work life and social life. Advances in machine learning and brain–computer interface (BCI) systems have pushed the ...

We incorporate effective components of the TasNet into a freq-domain separation method. We introduce a solution for directly optimizing the separation criterion in freq-domain networks. Our exp ...

Patients suffering from Parkinson's disease suffer from voice impairment. In this study, we introduce models to classify normal and Parkinson's patients using their speech. We used an AST (audio ...

Auditory stimulus reconstruction is a technique that finds the best approximation of the acoustic stimulus from the population of evoked neural activity. Reconstructing speech from the human auditory ...

Advanced Speech Emotion Detection (SED) using spectrogram analysis and deep learning offers a nuanced and accurate method for interpreting human emotions from speech. By transforming raw speech data ...