基于Kaldi的智能语音识别在物联网中的应用研究

Application Research of Kaldi-based Automatic Speech Recognition in the Internet of Things

智能语音识别技术是最常见的人类与机器交互的技术。随着物联网设备变得越来越常见,智能语音识别技术将在物联网中得到更广泛的应用。Kaldi作为目前较为优秀的跨平台智能语音识别套件,比较适合移植到物联网设备中。为此,采用煊扬G1为硬件实现平台,将智能风扇作为控制对象,将Kaldi套件移植到G1平台中。经过实验测试,Kaldi智能语音识别算法的识别度保持在85%以上,表明Kaldi套件可以应用在物联网设备中。由于硬件资源的限制,应该选择更适合的CPU核数、更多硬件平台来运行Kaldi框架,更好地保证智能语音识别的速度。

Automatic speech recognition technology is the most common interaction technology between humans and machines. As IoT devices become more common, automatic speech recognition technology will be more widely used in the IoT devices. As an excellent cross-platform automatic speech recognition suite, Kaldi is more suitable for porting to IoT devices. In this paper, G1 as Kaldi’s hardware implementation platform, and the automatic fan is used as the control object. After experimental tests, the recognition rate of Kaldi’s automatic speech recognition algorithm remains above 85%, indicating that Kaldi is more suitable for application in IoT devices. Limited by hardware resources, researchers should choose a more suitable hardware platform with more CPU cores to run the Kaldi framework, which can better ensure the speed of automatic speech recognition.