Method and device for processing speech based on artificial intelligence

A device for processing a speech based on artificial intelligence, comprising:

• a processor;
• and
• a memory, configured to store instructions executable by the processor, wherein the processor is configured to:
• perform a silence detection on a speech to be decoded;
• cut the speech to be decoded off to obtain a target speech if detecting that the speech to be detected is a silent speech;
• reset tail features of the target speech with preset tail features of silent frames;
• and
• perform a Connectionist Temporal Classification (CTC) decoding process on the target speech reset;
• wherein the processor is configured to reset tail features of the target speech with preset tail features of silent frames by acts of:
• selecting from the target speech, a plurality of tail frames of which features are to be reset;
• and
• resetting features of the plurality of tail frames with the preset tail features of silent frames;
• wherein the processor is configured to select from the target speech, a plurality of tail frames of which features are to be reset by acts of:
• determining a location of an utmost blank frame comprised in the target speech, wherein the utmost blank frame is a last blank frame of consecutive blank frames comprised in the target speech, and a first number that is the number of the consecutive blank frames is greater than a preset second number;
• and
• selecting a preset third number of frames backward from the location of the utmost blank frame as the tail frames.

A non-transitory computer readable storage medium comprising instructions, wherein when the instructions are executed by a processor of a device, the device is caused to perform a method for processing a speech based on artificial intelligence, and the method comprises:

• performing a silence detection on a speech to be decoded;
• cutting the speech to be decoded off to obtain a target speech if detecting that the speech to be detected is a silent speech;
• resetting tail features of the target speech with preset tail features of silent frames;
• and
• performing a Connectionist Temporal Classification (CTC) decoding process on the target speech reset;
• wherein resetting tail features of the target speech with preset tail features of silent frames comprises:
• selecting from the target speech, a plurality of tail frames of which features are to be reset;
• and
• resetting features of the plurality of tail frames with the preset tail features of silent frames;
• wherein selecting from the target speech, a plurality of tail frames of which features are to be reset comprises:
• determining a location of an utmost blank frame comprised in the target speech, wherein the utmost blank frame is a last blank frame of consecutive blank frames comprised in the target speech, and a first number that is the number of the consecutive blank frames is greater than a preset second number;
• and
• selecting a preset third number of frames backward from the location of the utmost blank frame as the tail frames.