Yih-Liang Shen | Searcharxiv

arXiv2024

Personalized Head-Related Transfer Function Prediction Based on Spatial Grouping

The head-related transfer function (HRTF) characterizes the frequency response of the sound traveling path between a specific location and the ear. When it comes to estimating HRTFs by neural network models, angle-specific models greatly outperform global models but demand high computational resources. To balance the computational resource and performance, we propose a method by grouping HRTF data spatially to reduce variance within each subspace. HRTF predicting neural network is then trained for each subspace. Simulation results show the proposed method performs better than global models and angle-specific models by using different grouping strategies at the ipsilateral and contralateral sides.

eess.SP↗

arXiv2018

Reinforcement Learning Based Speech Enhancement for Robust Speech Recognition

Conventional deep neural network (DNN)-based speech enhancement (SE) approaches aim to minimize the mean square error (MSE) between enhanced speech and clean reference. The MSE-optimized model may not directly improve the performance of an automatic speech recognition (ASR) system. If the target is to minimize the recognition error, the recognition results should be used to design the objective function for optimizing the SE model. However, the structure of an ASR system, which consists of multiple units, such as acoustic and language models, is usually complex and not differentiable. In this study, we proposed to adopt the reinforcement learning algorithm to optimize the SE model based on the recognition results. We evaluated the propsoed SE system on the Mandarin Chinese broadcast news corpus (MATBN). Experimental results demonstrate that the proposed method can effectively improve the ASR results with a notable 12.40% and 19.23% error rate reductions for signal to noise ratio at 0 dB and 5 dB conditions, respectively.

eess.AS↗