Google has published an update to the Lyra 0.0.2 audio codec, which is optimized to achieve maximum voice quality when using very slow communication channels. The codec was opened in early April, but was bundled with a proprietary math library. In version 0.0.2, this shortcoming was eliminated and an open replacement was created for the specified library - sparse_matmul, which, like the codec itself, is distributed under the Apache 2.0 license. Other improvements include the ability to use the Bazel build system with the GCC compiler and use this bundle by default on Linux instead of Bazel+Clang.
Recall that the quality of the transmitted voice data at low speeds Lyra is significantly superior to traditional codecs that use digital signal processing methods. In order to achieve high quality voice transmission in conditions of a limited amount of transmitted information, in addition to the usual methods of audio compression and signal conversion, Lyra uses a speech model based on a machine learning system that allows you to recreate the missing information based on typical speech characteristics. The model used to generate the sound has been trained using several thousand hours of voice recordings in over 70 languages. The performance of the proposed implementation is sufficient to encode and decode speech in real time on smartphones of the middle price range, with a signal transmission delay of 90 milliseconds.
Source: opennet.ru