Audio stream decoding method and device, storage medium and equipment

Abstract

The invention relates to the technical field of artificial intelligence. The invention discloses an audio stream decoding method and device, a storage medium and equipment. The method comprises the steps: firstly, acquiring an audio feature of a to-be-decoded audio stream, and performing first decoding on the audio feature to obtain a first intermediate decoding result, wherein the audio feature is obtained by performing feature extraction on the to-be-decoded audio stream, and the first decoding comprises resetting a system network state; and further determining a final decoding result of theto-be-decoded audio stream according to the first intermediate decoding result. Visibly, when the audio feature of the audio stream is subjected to the first decoding, the network state of the systemcan be reset, so that the state accumulation risk of the system can be eliminated, a final and more accurate decoding result can be conveniently obtained subsequently, the network delay of the wholedecoding process is low, the calculated amount is small, the power consumption is low, the state accumulation risk cannot be introduced, and the audio stream decoding accuracy is improved.

Description

technical field [0001] The present application relates to the technical field of artificial intelligence, and in particular to an audio stream decoding method, device, storage medium and equipment. Background technique [0002] With the rapid development of mobile Internet, Internet of Things, and artificial intelligence (AI) technologies, voice interaction, as a next-generation human-computer interaction method, has become an essential capability for almost all terminal devices. For example, mobile phones, tablet computers, watches, bracelets, etc. already have AI systems that implement voice interaction functions. Since the input signal of the AI ​​system is an audio stream, and as a serialized input signal, the audio stream has a strong correlation in both the time domain and the frequency domain. Therefore, in order to improve the processing accuracy of the audio stream, a recurrent neural network ( The recurrent neural network (RNN) structure is used as the network str...

AI Technical Summary

Problems solved by technology

[0003] At present, there are generally two methods for decoding audio streams: one is the streaming decoding method, although this method has the advantages of low network delay, small amount of calculation, and low power consumption, but when this method is used for decoding, the AI ​​system Every time the network performs inference operations at the current moment, it needs to use the system network status information at the previous moment, resulting in the inference operation results at the current moment relying on all the information from initialization to the current moment, which will introduce cumulative risk of the state; while the other The commonly used decoding method is the sliding window decoding method. Although this method does not introduce the risk of state accumulation, the disadvantages of this method are high network delay, large amount of calculation, and high power consumption.

Images