Summary of information:

Guangzhou Nine Chip Electronic Technology Co., Ltd. is a high -tech enterprise focusing on voice technology research. It mainly provides voice IC, voice chip, voice module, voice bips, voice alarm, forklift speeder, automotive speed limiter and other businesses Serve....

The voice recognition chip is also called voice recognition IC. Compared with the traditional voice chip, the characteristic of the large voice recognition chip is to be able to recognize voice. It allows the machine to understand human voice and can perform various actions according to the command. , Top mouth (smart doll). In addition, the voice recognition chip also has high -quality, high compression -recording recording function, which can realize human -machine dialogue.

The technologies involved in voice recognition chips include: signal processing, pattern recognition, probability theory and information theory, vocal mechanism and auditory mechanism, artificial intelligence, and so on.

The working principle of voice recognition chip

The embedded voice recognition system uses the principle of mode matching. The entered voice signal is first prepared, including the sampling of the voice signal, the anti -mixed filter, and the enhancement of the voice. The following is the feature extraction to extract a set of parameters that can describe the voice signal features from the voice signal waveform. The data after feature extraction is generally divided into two steps. The step is the system "learning" or "training" stage. The task of this stage is to build a reference mode library. Each word in the word table corresponds to a reference mode. It is based on this word. Repeat pronunciation multiple times, and then obtained in characterized extraction and some training. The second is the "identification" or "test" phase. According to certain criteria, the distortion measurement between the voice feature parameters and the voice information and the corresponding template in the mode library is matched. The matching is the identification result.

Classification of voice recognition chip

According to user restrictions, voice recognition chips can be divided into specific human voice recognition chips and non -specific human voice recognition chips.

Specific human voice recognition chip

The specific human voice recognition chip is aimed at the speech recognition of the designated person. If the words of others do not recognize, the user's voice reference sample must be stored in a database that is compared, that is, the voice training must be conducted before use. , Generally, you can use the voice entry for 2 times according to the machine prompts.

Non -specific human voice recognition chip

Non -specific human voice recognition is no need to recognize the identification technique of designated people, regardless of age and gender, as long as you speak the same language. The application mode is to collect about more than 200 people according to the determined more than a dozen voice interactive entries before the product is determined, and collect about 200 people. The sound sample is processed by the PC algorithm to obtain the voice model and characteristic database of the interactive entry, and then recorded it on the chip. Machines (smart dolls, electronic pets, and children's computers) applying this chip have interactive functions.

Non -specific human voice recognition applications are based on audioin -based algorithms. In this mode, there is no need to collect the sound samples of many people to do interactive recognition, but the disadvantage is that the recognition rate is not high and the recognition performance is unstable.

According to the continuity of the way of speaking, voice recognition chips can be divided into non -continuous voice recognition and continuous voice recognition.

Non -continuous voice recognition chip

For non -continuous voice, each word that recognizes must be identified separately, and it must be paused after each word is required.

Continuous voice recognition chip

Consecutive voice recognition can generally speak naturally and fluent to perform humanized voice recognition, but due to the connection between connectivity, it is difficult to achieve good identification effects.