云端AI与本地相结合的嵌入式语音识别系统

Local Embedded Speech Recognition System Combined with Cloud AI

综合利用了云端和本地语音识别算法,借助百度云端语音识别大数据库实现大量语音命令识别,和借助类似神经网络的深度学习算法使识别准确率提高,使语音特征数据库更新容易。结果表明,利用本地语音识别模组能准确响应特定词汇的语音识别,该方法方便快捷;并由STM32 MCU进行逻辑判断获取综合语音识别结果,提升了有效性和准确性。

System involves with cloud AI speech recognition platform,local integrated speech recognition module and mutual WiFi date transmission.The hardware parts of integrated speech recognition module are composed by SYN7318 module,ESP32 WiFi controller and STM32 MCU controller etc.The cloud and local speech recognition algorithms are utilized comprehensively.For one thing,the Baidu cloud speech recognition big data is used to achieve mass-speech command recognition.Further,the deep learning algorithm similar to neural network is adopted to improve recognition accuracy.The speech feature database can be updated easily.For another thing,the local speech recognition module is applied to respond speech recognition of specific vocabularies precisely,which is more convenient and efficient.The integrated speech recognition results obtained based on logical judgment of STM32 MCU improves efficiency and accuracy.