In the context of the intelligent age, we usually think that "software-defined hardware" is the mainstream of the times. But in many cases, the power of software is bounded, software can't do everything, and hardware is still the core of the future.
The working principle of the translator is that the built-in microphone recognizes the language and content of the user's speech, and converts the voice into text, and then translates the text into the text through the machine translation engine to convert the original language into the target language. Finally, the translated content is synthesized and played out.
Therefore, the translation machine is inseparable from the three core technologies: Automatic Speech Recognition, Machine Translation, and Speech Synthesis (or Text-to-Speech, TTS).
Among them, speech recognition is like "machine's auditory system", which allows the machine to convert speech signals into corresponding texts or commands through recognition and understanding. Since 2009, the deep learning research in the machine learning field has been introduced into the speech recognition acoustic model training, and the multi-layer neural network with RBM pre-training has been used to improve the accuracy of the acoustic model. At the same time, with the accumulation of big data corpus, speech recognition technology. Progress has been made by leaps and bounds.
As far as the use of the scene is concerned, when we simply use it once or twice, the translation software is enough to meet the needs of the vast majority of people. However, if you are traveling abroad or some high-frequency business scenarios, the translation software is far weaker than the translator because of its lack of hardware. Stability in the translation scenario refers to accuracy.
As far as the market is concerned, the relevant intelligent voice technology has been basically mature, and the reason why the hardware form of the translation machine is born is derived from the huge market demand.
In a noisy environment, the recognition rate of the phone by the mobile phone is not high; if the text is input and then translated, it is very time-consuming and not convenient. At the same time, the mobile phone as a personal item does not meet the attributes of the communication tool that the person communicates face to face. In a complex foreign network environment, mobile APP may not be able to achieve a good experience. Therefore, in the real environment, software and hardware integration products are a better solution