CPqD Speech Recognition
The most recent technologies to allow conversations between people and any system using real-time speech recognition. Simple, flexible and reliable.
CPqD Speech Recognition is an online resource that recognizes speech in Portuguese with extreme accuracy. The system was trained with the speech of thousands of Brazilians with accents from all over Brazil, resulting in highly-efficient and engaging automated customer service applications.
The speech recognition engine uses advanced deep learning neural network algorithms and is able to recognize application audio with highest level of accuracy.
Processes audio as it is received and returns partial results while the speech is still being recognized. Besides online recognition, the system returns text based on the recognition of prerecorded linear PCM audio, with sample rates of 8 kHz or 16kHz.
The system is accurate in different noisy environments, with no need for previous audio signal processing or noise cancellation.
CPqD Speech Recognition supports any application or device that can be integrated via HTTP REST, MRCP V1 and V2 or Websocket interfaces.
Software development kits are available in C++ and Java.
The product supports two modes: free speech, for generic scenarios where there is no need to create grammars; and grammar mode, with grammar written in the SRGS/SISR standard as a form of customization for a specific use context.
The system strips silence at the beginning and ending of audio phrases. This avoids unneeded processing to recognize something that is not speech.
Talk to one of our specialists.
CPqD develops and offers robust network analysis and structuring projects with operational and energetic efficiency, as well as predictive maintenance.Talk to a Specialist