摘要
本文研究的是低误触发高识别率离线语音识别冰箱及控制系统。以单片机为处理中心,以预先约定的音频格式通过ADC采集录音,将采集到的音频数据结合算法来实现语音识别。以降噪算法结合三MIC算法的语音识别系统作为冰箱人机交互装置,加入人体检测和唤醒词触发,并结合MCU滤波算法和信号加窗处理极大降低了周期性环境噪声对语音交互的干扰,降低了语音交互的误触发率,并提高了语音识别的准确率。加入本方案语音识别控制的冰箱具有很强的实用性,给用户操作带来很大的方便,在智能冰箱人机交互上是一个很大提升,且成本低、系统简洁,是继冰箱触摸按键操作之后的一种具有突破性操作方式,大大提高了用户体验,从而增强了产品的市场竞争力,具有潜在的开发拓展前景。
This paper locuses on the low error rate, high recognition rate, off-line speech recognition reffigeralor and control system. Taking MCU as the processiug center, the audio data is collecled by ADC in adwmce, and the collected audio data are combined with the algorithm to realize speech recognition. The speech recognition system with the directional noise reduction algorithm combined with the three MIC noise reduction algorithm is used as the human-computer interaction device of the refrigerator, adding the human detection and the arousal word trigger, and combining the MCU fihering algorithm and the signal window processing to greatly reduce the interference of the periodic environment noise to the speech interaction, and reduce the false trigger rate of the speech interaction, and improve the speech interaction. The accuracy of speech recognition. The refrigerator with the speech recognitinn control of this scheme is very practical, brings great couvenience to the operation of the user. h is a great promotion, low cost and simple system in the human computer interaction of the intelligent refrigerator. It is a breakthrough operation method after the touch button operation of the refi-igenltor. which greatly improves the user experience. Thus, the market competitiveness of products is enhanced and potential development and expansion prospects are possible.
作者
罗莉文
陈庆传
刘道金
韩斌斌
LUO Liwen;CHEN Qingchuan;LIU Daojin;HAN Binbin(Hangzhou Huari home appliance Co.,Ltd.Hangzhou 311401)
出处
《家电科技》
2018年第9期55-57,61,共4页
Journal of Appliance Science & Technology
关键词
语音识别
智能控制
降噪算法
三麦阵列
低误触发
人体检测
阀值滤波
Speech recognition
Intelligent control
Directional noise reduction
Three wheat array
Low falsetriggering
Human detection
Threshold filtering