System and method for extracting and using prosody features
A system for carrying out voice pattern recognition and a method for achieving same. The system includes an arrangement for acquiring an input voice, a signal processing library for extracting acoustic and prosodic features of the acquired voice, a database for storing a recognition dictionary, at least one instance of a prosody detector for carrying out a prosody detection process on extracted respective prosodic features, communicating with an end user application for applying control thereto.
1. A method for applying voice pattern recognition, implementable on an input voice, said method comprising the steps of:
acquiring said input voice
extracting prosodic features from said input voice at least once; and
carrying out a voice pattern classification process using dynamic time warping by integrating pattern matching with said extracted prosodic features to improve recognition performance using a Sakoe-Chuba search space, and reducing thereby the size of the search space on the basis of a detected pattern and a predetermined respective database entry in order to produce an output of said voice pattern classification process.
2. The method of claim 1 , in which a 3 rd party speech pattern recognition engine is used for providing a recognized pattern as an input to an end user application.
3. An automated assistant for speech disabled people, operating on a computing device, said assistant comprising:
an input device for receiving user input voice wherein the input device comprises at least a speech input device for acquiring voice of said people;
a signal library for extracting acoustic and prosodic features of said input voice;
at least one prosody detector for extracting respective prosodic features;
a database for storing a recognition dictionary based on predetermined mapping between voice features extracted from voice recording from said people and a reference;
a voice pattern classifier in which one of said at least one prosody detector is integrated in dynamic time warping by integrating pattern matching with said extracted prosodic features to improve recognition performance using a Sakoe-Chuba search space, and reducing thereby the size of the search space for rendering an output; and
an output device, for rendering said output.