Following the alphago Google deepmind in the voice of a step on the synthesis 下北glory days

After AlphaGo of Google DeepMind in sound synthesis a further step on the original title: following the new achievements of Google AlphaGo Deepmind step towards a breakthrough of the car under the Google Deepmind announced on Iran male computer synthesized speech has taken an important step in sound synthesis. According to FT reports, DeepMind released the latest research results WaveNet makes the difference between computer-generated speech and human voice to reduce the difference in half. Deepmind wrote in his blog, "let the human and machine dialogue is a long dream in the field of human-computer interaction". Google Deepmind Bloomberg reported in English and Chinese blind test, the listener generally believe that Google compared to any existing use different technology text to speech conversion program, WaveNet generates more natural speech. According to company related personnel, WaveNet has not been applied to any Google products. The system requires a strong computing power to synthesize speech, so it should not be used in the near future in the "real scene" to go. Google DeepmindDeepMind is a focus on artificial intelligence company, founded in 2010 in London, and in 2014 by Google to 400 million pounds (about $532 million) acquisition. The technology used by WaveNet is fundamentally different from the existing artificial sound synthesizer. WaveNet is the actual sound waves, not the language itself, and the text is converted into sound. The neural network is used to simulate the human brain. This comes from the support of a powerful database of up to 16000 sample waveforms per second. In addition, each sample should be based on the previous sample to establish the prediction of the acoustic pattern. DeepMind staff said, this is indeed a very challenging task." Many of the existing artificial sound synthesizer is to record a large number of human voice fragments, to build a large database, and to re combine these fragments into a complete discourse. The sound produced is close to the true voice, but not completely natural. The disadvantage is that the resulting sound cannot be changed simply. According to DeepMind, other systems can generate sounds electronically, so that the sound can be modified simply, but not close to the real voice. Sound synthesis in artificial intelligence, although people seem to have been more interested in speech recognition, speech synthesis and a computer this neglected area. But the major technology companies do not let go of the application of more and more common human-computer interaction. Amazon, apple, Microsoft and Google’s parent company Alphabet have put personal digital assistants interact with customers. Earlier this year, DeepMind’s artificial intelligence AlphaGo beat South Korea go master Li Shishi, has aroused widespread concern. Google has been on how to use the technology of artificial intelligence called tight lipped, have been using these techniques to reduce the data center 40% of the electricity, left.相关的主题文章: