论文部分内容阅读
大词汇量连续语音识别由于其庞大的计算量和对存储单元巨大的需求 ,一直难以实时实现。但随着 DSP技术的发展 ,对于中小词汇量连续语音的实时实现是有可能的。文中尝试采用 TI公司新型号的 DSP芯片 ,实现属于中小词汇量连续语音的数字连接词的实时识别。经实验表明 ,该系统基本能实现预期识别目标
Large vocabulary continuous speech recognition has been difficult to achieve in real time due to its huge amount of computation and huge demand for storage units. However, with the development of DSP technology, it is possible to realize the real-time continuous speech of small and medium vocabulary. The paper attempts to use TI’s new model of DSP chip, which realizes the real-time identification of digital connectives that belong to the continuous pronunciation of small and medium vocabulary. The experiment shows that the system can basically achieve the expected target