I'm trying to implement a server capable of doing speech recognition. But this last part has to be implemented through a C\C++ Addon. Right now, I'm saving the wav file on the server and passing the name to the addon:
var obj1 = julius("nomeficheiro.wav","lista|palavras|possiveis"); word = obj1.recog;//resultado cm = obj1.cm;//confiança no resultado
However my intention is: not write files to disk and do everything through memory. The speech recognizer accepts stdin as possible input. Is it possible to pipe between the C\C++ addon and the NodeJs?
Yes, it is possible. Node.js has an infrastructure for running processes. This infrastructure is gathered in the Child Process module.
In node.js you can perform the piping procedure with a child process by supplying a readable stream as stdin. For more documentation, visit the subprocess.stdin section of the official documentation.