Abstract: A processing system receives an audio signal encoding an utterance and determines that a first portion of the audio signal corresponds to a predefined phrase. The processing system accesses one or more text-dependent models associated with the predefined phrase and determines a first confidence based on the one or more text-dependent models associated with the predefined phrase, the first confidence corresponding to a first likelihood that a particular speaker spoke the utterance. The processing system determines a second confidence for a second portion of the audio signal using one or more text-independent models, the second confidence corresponding to a second likelihood that the particular speaker spoke the utterance. The processing system then determines that the particular speaker spoke the utterance based at least in part on the first confidence and the second confidence.

Abstract: The voiced sound interval classification device comprises a vector calculation unit which calculates, from a power spectrum time series of voice signals, a multidimensional vector series as a vector series of a power spectrum having as many dimensions as the number of microphones, a difference calculation unit which calculates, with respect to each time of the multidimensional vector series, a vector of a difference between the time and the preceding time, a sound source direction estimation unit which estimates, as a sound source direction, a main component of the differential vector, and a voiced sound interval determination unit which determines whether each sound source direction is in a voiced sound interval or a voiceless sound interval by using a predetermined voiced sound index indicative of a likelihood of a voiced sound interval of the voice signal applied at each time.

Abstract: Octave errors may be reduced during pitch determination for noisy audio signals. Pitch may be tracked over time by determining amplitudes at harmonics for individual time windows of an input signal. Octave errors may be reduced in individual time windows by fitting amplitudes of corresponding harmonics across successive time windows to identify spurious harmonics caused by octave error. A given harmonic may be identified as either being associated with the same pitch as adjacent harmonics in the given time window or being spurious based on parameters of the fitting function.

Abstract: Voice enhancement and/or speech features extraction may be performed on noisy audio signals using successively refined transforms. Downsampled versions of an input signal may be obtained, which include a first downsampled signal with a lower sampling rate than a second downsampled signal. Successive transforms may be performed on the input signal to obtain a corresponding sound model of the input signal. The successive transforms performed may include: (1) performing a first transform on the first downsampled signal to yield a first pitch estimate; (2) performing a second transform on the second downsampled signal to yield a second pitch estimate and a first harmonics estimate based on the first pitch estimate; and (3) performing a third transform on the input signal to yield a third pitch estimate and a second harmonics estimate based on the second pitch estimate and the first harmonics estimate.

Abstract: Techniques are described for translating natural language input to a machine-readable form that accurately represents the semantic meaning of the input intended by the user.

Abstract: Techniques for use in connection with a system for automatically generating text. Techniques include accessing information specifying at least one referential expression for at least a first referent and at least one anaphoric expression for at least the first referent; accessing a template that includes human-language text and a first tag that serves as a placeholder for a first text portion including a reference to at least the first referent; automatically identifying, using at least one system rule and at least one processor, text to use for the first text portion at least in part by determining whether to use as the text for the first text portion the at least one referential expression or the at least one anaphoric expression; and automatically generating output text including the human-language text and the identified text for the first text portion.

Abstract: A log message is processed. The log message to be processed is received. One or more portions of the log message to be separately extracted are identified. A value is extracted from each identified portion. Extracting the value includes using an extraction rule. The extraction rule is associated with the identified portion.

Abstract: A speech processing apparatus acquires pseudo speech from a mixture of sound including desired speech and noise. A first microphone inputs a first mixture sound, including desired speech and noise, and outputs a first mixture signal. A second microphone opens to the sound space and is disposed at a focus position of an interface, that is part of a boundary of the sound space and has one of a quadratic surface shape and a pseudo surface shape approximating a quadratic surface, inputs a second mixture sound including the desired speech reflected by the interface and the noise reflected by the interface at a ratio different from the first mixture sound, and outputs a second mixture signal. A noise suppression circuit suppresses an estimated noise signal based on the first mixture signal and the second mixture signal and outputs a pseudo speech signal.

Abstract: A voice signal processing method processes voice signals acquired by a microphone. A voice processing device acquires first voice signals according to a first sampling frequency, and samples second voice signals from the first voice signals according to a second sampling frequency. The second voice signals are encoded to obtain a basic voice package. A voiceprint data package of each voice signal frame of the first voice signals is obtained using a curve fitting method, and a pitch data package of each voice signal frame of the first voice signals is obtained according to pitch distribution of twelve central octave keys of a standard piano. The voiceprint data package and the pitch data package are embedded into the basic audio package to generate a final voice package of the first voice signals.