Natural human-computer voice interaction method and system

Abstract

The invention discloses a natural human-computer voice interaction method. The method comprises: (1) signal acquisition is carried out; to be specific, a microphone array is used for collecting sound signals of a sound source, the orientation of the sound source is determined based on the sound signals collected by microphones at different places in the microphone array, and a camera is used for collecting an image signal of the orientation of the sound source; (2), sound source identity identification is carried out; to be specific, according to the collected image signal, the identity of the sound source is identified; (3), speaking object identification is carried out; to be specific, according to the collected image signal and sound signals, a speaking object is identified based on a hot word monitoring and slight line analysis algorithm; and (4), on the basis of the collected sound signals, whether a response to the sound source needs to be made is determined. Correspondingly, the invention also discloses a natural human-computer voice interaction system. The natural human-computer voice interaction method and system can be applied to human-computer voice interaction to enhance the human-computer voice interaction experience, so that the user can obtain a natural feeling during the human-computer voice interaction process.

Description

technical field [0001] The present invention relates to a human-computer voice interaction method and system, in particular to a natural human-computer voice interaction method and system. Background technique [0002] With the rapid development of natural language processing technologies such as speech recognition and semantic understanding in recent years, digital intelligent assistants with speech as the main interaction mode are emerging, such as Apple Siri, Google Now, Microsoft Cortana, Amazon Echo, etc. As a way of human-computer interaction, voice is widely used in terminals such as smart watches, smart phones, tablet computers, and personal computers. [0003] The use process of current mainstream voice assistants usually has the following characteristics: [0004] 1. Usually, the working state of an intelligent voice assistant on a terminal is divided into a standby state and a recognition state. [0005] 2. The intelligent voice assistant in the standby state wi...

AI Technical Summary

Problems solved by technology

[0007] One of the main reasons for the design of the existing mainstream systems is that from speech to text recognition and semantic understanding is still a matter that consumes quite a lot of computing resources, so most of the solutions are based on the cloud. computing resources, there are also huge privacy concerns

[0008] However, there are many limitations in this voice interaction method, which makes users feel unnatural during use, for example:

[0009] 1. After entering the recognition state, everything the user said will be used as instructions, and it does not have the ability to distinguish whether a piece of voice is requested by the user, and it is easy to record error messages

[0010] 2. Limited by the current hardware and software performance, it is impossible to have the ability of accurate speaker recognition only by relying on voice information, that is to say, a single terminal cannot handle the requests of multiple users

[0011] 3. The switching process of the working state is unnatural, especially in the pure voice scene, the user needs to speak the specified vocabulary to wake up the voice assistant

When multiple devices within the same range are equipped with voice assistant software, they may even be woken up at the same time

Images