Wake-up word recognition training system and method
A wake-up word recognition training system includes: a sentence database, storing a plurality of sentences and a phoneme sequence and a speech signal corresponding to each of the sentences; a phoneme disassembly module, disassembling a wake-up word inputted from the outside to obtain a wake-up word phoneme sequence; a phoneme analysis module, matching the wake-up word phoneme sequence to the sentences and/or phoneme sequences thereof, to obtain wake-up word part-of-speech sentences and non-wake-up word part-of-speech sentences; a sentence classification module, dividing the sentences in the sentence database into the wake-up word part-of-speech sentences and the non-wake-up word part-of-speech sentences according to a comparison result of the phoneme comparison module; and a wake-up word recognition module, obtaining speech signal fragments of the wake-up word and a non-wake-up word according to a phoneme combination of the wake-up word part-of-speech sentences and the non-wake-up word part-of-speech sentences.
1 . A wake-up word recognition training system, comprising:
a sentence database, storing a plurality of sentences and a phoneme sequence and a speech signal corresponding to each of the sentences;
a phoneme disassembly module, disassembling a wake-up word inputted from the outside to obtain a wake-up word phoneme sequence;
a phoneme analysis module, matching the wake-up word phoneme sequence to the sentences and/or phoneme sequences thereof, to obtain wake-up word part-of-speech sentences and non-wake-up word part-of-speech sentences; and
a wake-up word recognition module, constructing a deep neural network model according to a phoneme combination of the wake-up word part-of-speech sentences and the non-wake-up word part-of-speech sentences, and training the deep neural network model according to phoneme sequences and speech signals of the wake-up word part-of-speech sentences and the non-wake-up word part-of-speech sentences, to establish a wake-up word recognition model;
wherein the wake-up word recognition training system includes an input module, a sentence database, and a processing module, where the processing module includes a phoneme disassembly module, a phoneme comparison module, a sentence classification module, a wake-up word recognition module and a generic speech recognition module; and
wherein text input or at least one of a speech input device and a microphone is used to input a personalized wake-up word to obtain the wake-up word.
2 . The wake-up word recognition training system according to claim 1 , wherein the phoneme analysis module comprises a phoneme comparison module and a sentence classification module; the phoneme comparison module compares the sentences and the phoneme sequences in the sentence database according to the wake-up word phoneme sequence; and the sentence classification module divides the sentences in the sentence database into the wake-up word part-of-speech sentences and the non-wake-up word part-of-speech sentences according to a comparison result of the phoneme comparison module.
3 . The wake-up word recognition training system according to claim 1 , wherein constituent units of the phoneme sequence comprise a sequence of three consecutive phonemes, a sequence of two consecutive phonemes, and/or a single phoneme.
4 . The wake-up word recognition training system according to claim 1 , wherein phoneme sequences and speech signal fragments corresponding to the sentences of the wake-up word in the wake-up word part-of-speech sentences of the sentence database are found out to serve as training data about the wake-up word for the deep neural network model by the wake-up word recognition module using a generic speech recognition module and a dynamic programming algorithm.
5 . The wake-up word recognition training system according to claim 4 , wherein sentences corresponding to a non-wake-up word and/or speech signal fragments of phoneme sequences thereof in the sentence database are found out to serve as training data about the non-wake-up word for the deep neural network model by the wake-up word recognition module using a method of text comparison.
6 . A wake-up word recognition training method, comprising
disassembling, by using a phoneme disassembly module, a wake-up word inputted from the outside to obtain a wake-up word phoneme sequence;
matching, by using a phoneme analysis module, the wake-up word phoneme sequence to each of a plurality of sentences and/or phoneme sequences corresponding to the sentences, to obtain wake-up word part-of-speech sentences and non-wake-up word part-of-speech sentences; and
constructing, by using a wake-up word recognition module, a deep neural network model according to a phoneme combination of the wake-up word part-of-speech sentences and the non-wake-up word part-of-speech sentences, and training the deep neural network model according to phoneme sequences and speech signals of the wake-up word part-of-speech sentences and the non-wake-up word part-of-speech sentences, to establish a wake-up word recognition model;
wherein the wake-up word recognition training system includes an input module, a sentence database, and a processing module, where the processing module includes a phoneme disassembly module, a phoneme comparison module, a sentence classification module, a wake-up word recognition module and a generic speech recognition module; and
wherein text input or at least one of a speech input device and a microphone is used to input a personalized wake-up word to obtain the wake-up word.
7 . The wake-up word recognition training method according to claim 6 , wherein constituent units of the phoneme sequence comprise a sequence of three consecutive phonemes, a sequence of two consecutive phonemes, and/or a single phoneme.
8 . The wake-up word recognition training method according to claim 6 , wherein phoneme sequences and speech signal fragments corresponding to the sentences of the wake-up word in the wake-up word part-of-speech sentences of the sentence database are found out to serve as training data about the wake-up word for the deep neural network model by the wake-up word recognition module using a generic speech recognition module and a dynamic programming algorithm.
9 . The wake-up word recognition training method according to claim 8 , wherein sentences corresponding to a non-wake-up word and/or speech signal fragments of phoneme sequences thereof in the sentence database are found out to serve as training data about the non-wake-up word for the deep neural network model by the wake-up word recognition module using a method of text comparison.
10 . The wake-up word recognition training method according to claim 6 , wherein the matching, by using a phoneme analysis module, the wake-up word phoneme sequence to each of a plurality of sentences and/or phoneme sequences corresponding to the sentences, to obtain wake-up word part-of-speech sentences and non-wake-up word part-of-speech sentences comprises: comparing, by using a phoneme comparison module, the sentences and the phoneme sequences in the sentence database according to the wake-up word phoneme sequence; and obtaining, by using a sentence classification module, the wake-up word part-of-speech sentences and the non-wake-up word part-of-speech sentences from the sentences according to a comparison result of the phoneme comparison module.