一种高性能固定语音识别并行处理架构

High-performance parallel architecture for specific speech recognition

针对固定语音识别系统后端大量的模板匹配识别运算,提出了一种多处理单元并行超标量匹配处理架构,有效地提升了大容量模板库条件下实时处理性能。将所提出的并行匹配处理架构在Virtex-6 FPGA上进行了实现,其结果显示,在保证识别精度的情况下,处理能力得到显著提高。

Aiming at the large quantity computation of matching and recognition in the back-end, this paper proposed a paral- lel architecture of multiple-PE（ processing element）which achieved effective real-time processing performance. The proposed architecture has been implemented on the Virtex-6 FPGA, which significantly improved the processing performance in the ease of ensuring the accuracy of recognition.