Detection and enhancement of speech in binaural recordings

Disclosed herein are methods, systems, and computer program products for segmenting a binaural recording of speech into portions containing own speech and portions containing external speech, and processing each category using different settings to obtain an enhanced overall presentation. Segmentation is performed based on a combination of i) feature-based frame-by-frame classification, and ii) detection of dissimilarity by a statistical method. The segment information is then used by a speech enhancement chain, with independent settings for processing the own speech portion and the external speech portion. 本文公开了用于将语音的双耳录音分段成包含自身语音的部分和包含外部语音的部分，并使用不同设置处理每个种类以获得增强的整体呈现的方法、系统和计算机程序产品。基于以下组合进行分段：i)基于特征的逐帧分类，和ii)通过统计方法检测相异度。分段信息随后被语音增强链使用，其中独立设置用于处理自身语音部分和外部语音部分。