Design and Construction of Intelligent
Voice Control System
Dapeng Zhu
(B) , Shu Liu, and Chao Liu
Letter Box 21, Urumqi 83001, China
lingo renlo@qq.com
Abstract. With the development of speech recognition technology,
there are wide applications of intelligent voice around the world. In the
paper, we propose an intelligent voice control system which can realize the functions of voice activity detection, speech recognition, speaker
authentication, instruction analysis, and automatic answer. A new voice
activity detection (VAD) algorithm and backtracking algorithm are proposed to improve system performance. Then, we test our system in our
own silumation corpus, we create the system evaluation indexes and the
text result is satisfactory.
Keywords: Speech recognition · VAD · Backtracking algorithm
1 Introduction
Speaker recognition (recognizing who was speaking) and speech recognition (recognizing what people said) are used widely in the field of artificial intelligence.
For example, paper [1] designs a specific human voice recognition home control
system by improving the corresponding algorithm. Paper [2] designs a Chinese
text and audio editing system based on AI, improving the editing efficiency of
audio content and the timeliness of audio news release.
In this paper, we propose an intelligent voice control system. The system
uses speech recognition and speaker recognition technologies based on deep neural network. It can recognize specific speakers and process his or her continuous voice command in real time, realizing deep neural network interference
and instruction analysis. We do not use the online services of speech-to-text
transcription and voiceprint authentication supported by Internet providers. We
build specialized instruction dataset and relevant Chinese corpus. In consideration of multi-speakers, high background noise, random speaking voice pauses,
and no wake-up-word for speech recognition, we propose voice activity detection
algorithm and backtracking algorithm to recognize right instruction. In addition,
we apply Chinese initial and tunal final speech model label method. Finally, we
define the system evaluation indexes to show us the system performance from
multi-aspects.
c
Springer Nature Singapore Pte Ltd. 2020
Q. Liang et al. (Eds.): Artificial Intelligence in China, LNEE 572, pp. 319–325, 2020.
https://doi.org/10.1007/978-981-15-0187-6_37
Voice Control System
Dapeng Zhu
(B) , Shu Liu, and Chao Liu
Letter Box 21, Urumqi 83001, China
lingo renlo@qq.com
Abstract. With the development of speech recognition technology,
there are wide applications of intelligent voice around the world. In the
paper, we propose an intelligent voice control system which can realize the functions of voice activity detection, speech recognition, speaker
authentication, instruction analysis, and automatic answer. A new voice
activity detection (VAD) algorithm and backtracking algorithm are proposed to improve system performance. Then, we test our system in our
own silumation corpus, we create the system evaluation indexes and the
text result is satisfactory.
Keywords: Speech recognition · VAD · Backtracking algorithm
1 Introduction
Speaker recognition (recognizing who was speaking) and speech recognition (recognizing what people said) are used widely in the field of artificial intelligence.
For example, paper [1] designs a specific human voice recognition home control
system by improving the corresponding algorithm. Paper [2] designs a Chinese
text and audio editing system based on AI, improving the editing efficiency of
audio content and the timeliness of audio news release.
In this paper, we propose an intelligent voice control system. The system
uses speech recognition and speaker recognition technologies based on deep neural network. It can recognize specific speakers and process his or her continuous voice command in real time, realizing deep neural network interference
and instruction analysis. We do not use the online services of speech-to-text
transcription and voiceprint authentication supported by Internet providers. We
build specialized instruction dataset and relevant Chinese corpus. In consideration of multi-speakers, high background noise, random speaking voice pauses,
and no wake-up-word for speech recognition, we propose voice activity detection
algorithm and backtracking algorithm to recognize right instruction. In addition,
we apply Chinese initial and tunal final speech model label method. Finally, we
define the system evaluation indexes to show us the system performance from
multi-aspects.
c
Springer Nature Singapore Pte Ltd. 2020
Q. Liang et al. (Eds.): Artificial Intelligence in China, LNEE 572, pp. 319–325, 2020.
https://doi.org/10.1007/978-981-15-0187-6_37