The Cainteoir Engine is a library for reading and recording different document formats (ePub, HTML, MHT, RTF, email, and others) to various audio output formats (such as PulseAudio, WAV, and Ogg/Vorbis). It also provides the following command-line tools: cainteoir, a front-end to the Cainteoir text-to-speech library; metadata, which extracts metadata from documents to RDF tuples; and tagcloud, which generates tag clouds and tag cloud data.
libneuro-java is a Java native interface to libneuro. It provides C++ and Java wrappers, as well as Java classes. It also provides control files to generate them from the libneuro interface using the program swig. The major and minor versions of this JNI will track that of libneuro, but its bugfixes are independent of libneuro bugfixes.
libalinga-java is a Java native interface to libalinga. It provides C++ and Java wrappers, as well as Java classes. It also provides control files to generate them from the libalinga interface using the program swig. The major and minor versions of this JNI will track that of libalinga, but its bugfixes are independent of libalinga bugfixes.
libalinga is a C++ implementation of a multi-stream codec for the ALingA (Aligned Linguistic Annotation) format. It makes use of libogg++. Each ALingA stream holds at least one stream of annotation data, which is in the LingA format. It may also interleave the signal stream(s) against which the LingA streams are aligned, or it may simply reference such streams. It also provides metadata about the underlying manifold for the signals and the annotations. The metadata is ordered for runtime parsing of the number and type of signal and LingA codecs to enable decoding of the multiple logical streams in one pass.
PED is a dialogue management system that uses a probabilistic nested belief model to choose dialogue strategies. The dialogue system designer need only supply a set of plan rules to PED as a dialogue grammar with preconditions. Using this grammar, PED constructs game trees (like the one below) to represent the outcomes of the dialogue, so that a dialogue strategy can be chosen for the current turn in the dialogue. PED automatically maintains a belief model by a belief revision process that uses the observed acts in the dialogue. The game tree is evaluated in the context of this belief model. PED is efficient because it uses probabilistic estimates of belief rather than a plain logical belief model.
Intelligent systems would benefit from being able to localize and track sound sources in real life settings. Such a capability can help in localizing a person or an interesting event in the environment, and also provides enhanced processing for other capabilities such as speech recognition. To give this capability to a computing system, the challenge is not only to localize simultaneous sound sources, but to track them over time. The ManyEars project propose a robust sound source localization and tracking method using an array of eight microphones. The method is based on a frequency-domain implementation of a steered beamformer along with a particle filter-based tracking algorithm. Tests on a mobile robot show that the algorithm can localize and track in real-time multiple moving sources of different types over a range of 7 meters. These new capabilities allowed the robot to interact using more natural means with people in real life settings. The ManyEars project provides an easy to use C library for microphone array processing. This includes sound source localisation, tracking, and separation.A tuning Qt GUI is also available for fine tuning the parameters.