Abstract: Solutions for authenticating digital audio include: generating a first band-limited watermark using a first key, generating a second band-limited watermark using a second key, wherein the bandwidth of the second watermark does not overlap with the bandwidth of the first watermark; and embedding the first watermark and the second watermark into a segment of the digital audio file. Solutions also include determining a first watermark score of a segment of the digital audio file for the first watermark using the first key; determining a second watermark score of the segment of the digital audio file for the second watermark using the second key; based on at least the first watermark score and the second watermark score, determining a probability that the digital audio file is watermarked; and generating a report indicating whether the digital audio file is watermarked. In some examples, solutions may also embed and decode messages.

Abstract: The text-to-speech audio HIP technique described herein in some embodiments uses different correlated or uncorrelated words or sentences generated via a text-to-speech engine as audio HIP challenges. The technique can apply different effects in the text-to-speech synthesizer speaking a sentence to be used as a HIP challenge string. The different effects can include, for example, spectral frequency warping; vowel duration warping; background addition; echo addition; and varying the time duration between words, among others. In some embodiments the technique varies the set of parameters to prevent using Automated Speech Recognition tools from using previously used audio HIP challenges to learn a model which can then be used to recognize future audio HIP challenges generated by the technique. Additionally, in some embodiments the technique introduces the requirement of semantic understanding in HIP challenges.

Abstract: Embodiments of an audio-to-video engine are disclosed. In operation, the audio-to-video engine generates facial movement (e.g., a virtual talking head) based on an input speech. The audio-to-video engine receives the input speech and recognizes the input speech as a source feature vector. The audio-to-video engine then determines a Maximum A Posterior (MAP) mixture sequence based on the source feature vector. The MAP mixture sequence may be a function of a refined Gaussian Mixture Model (GMM). The audio-to-video engine may then use the MAP to estimate video feature parameters. The video feature parameters are then interpreted as facial movement. The facial movement may be stored as data to a storage module and/or it may be displayed as video to a display device.

Abstract: An automated method of providing a pronunciation of a word to a remote device is disclosed. The method includes receiving an input indicative of the word to be pronounced. The method further includes searching a database having a plurality of records. Each of the records has an indication of a textual representation and an associated indication of an audible representation. At least one output is provided to the remote device of an audible representation of the word to be pronounced.

Abstract: Instead of relying on humans to subjectively evaluate speech intelligibility of a subject, a system objectively evaluates the speech intelligibility. The system receives speech input and calculates confidence scores at multiple different levels using a Template Constrained Generalized Posterior Probability algorithm. One or multiple intelligibility classifiers are utilized to classify the desired entities on an intelligibility scale. A specific intelligibility classifier utilizes features such as the various confidence scores. The scale of the intelligibility classification can be adjusted to suit the application scenario. Based on the confidence score distributions and the intelligibility classification results at multiple levels an overall objective intelligibility score is calculated. The objective intelligibility scores can be used to rank different subjects or systems being assessed according to their intelligibility levels. The speech that is below a predetermined intelligibility (e.g.

Abstract: Frame mapping-based cross-lingual voice transformation may transform a target speech corpus in a particular language into a transformed target speech corpus that remains recognizable, and has the voice characteristics of a target speaker that provided the target speech corpus. A formant-based frequency warping is performed on the fundamental frequencies and the linear predictive coding (LPC) spectrums of source speech waveforms in a first language to produce transformed fundamental frequencies and transformed LPC spectrums. The transformed fundamental frequencies and the transformed LPC spectrums are then used to generate warped parameter trajectories. The warped parameter trajectories are further used to transform the target speech waveforms in the second language to produce transformed target speech waveform with voice characteristics of the first language that nevertheless retain at least some voice characteristics of the target speaker.

Abstract: Described is a technology by which synthesized speech generated from text is evaluated against a prosody model (trained offline) to determine whether the speech will sound unnatural. If so, the speech is regenerated with modified data. The evaluation and regeneration may be iterative until deemed natural sounding. For example, text is built into a lattice that is then (e.g., Viterbi) searched to find a best path. The sections (e.g., units) of data on the path are evaluated via a prosody model. If the evaluation deems a section to correspond to unnatural prosody, that section is replaced, e.g., by modifying/pruning the lattice and re-performing the search. Replacement may be iterative until all sections pass the evaluation. Unnatural prosody detection may be biased such that during evaluation, unnatural prosody is falsely detected at a higher rate relative to a rate at which unnatural prosody is missed.

Abstract: An exemplary method includes receiving stroke information for a partially written East Asian character, the East Asian character representable by one or more radicals; based on the stroke information, selecting a radical on a prefix tree wherein the prefix tree branches to East Asian characters as end states; identifying one or more East Asian characters as end states that correspond to the selected radical for the partially written East Asian character; and receiving user input to verify that one of the identified one or more East Asian characters is the end state for the partially written East Asian character. In such a method, the selection of a radical can occur using radical-based hidden Markov models. Various other exemplary methods, devices, systems, etc., are also disclosed.

Abstract: The text-to-speech audio HIP technique described herein in some embodiments uses different correlated or uncorrelated words or sentences generated via a text-to-speech engine as audio HIP challenges. The technique can apply different effects in the text-to-speech synthesizer speaking a sentence to be used as a HIP challenge string. The different effects can include, for example, spectral frequency warping; vowel duration warping; background addition; echo addition; and varying the time duration between words, among others. In some embodiments the technique varies the set of parameters to prevent using Automated Speech Recognition tools from using previously used audio HIP challenges to learn a model which can then be used to recognize future audio HIP challenges generated by the technique. Additionally, in some embodiments the technique introduces the requirement of semantic understanding in HIP challenges.

Abstract: A representation of a speech signal is received and is decoded to identify a sequence of position-dependent phonetic tokens wherein each token comprises a phone and a position indicator that indicates the position of the phone within a syllable.

Abstract: Embodiments of rich context modeling for speech synthesis are disclosed. In operation, a text-to-speech engine refines a plurality of rich context models based on decision tree-tied Hidden Markov Models (HMMs) to produce a plurality of refined rich context models. The text-to-speech engine then generates synthesized speech for an input text based at least on some of the plurality of refined rich context models.

Abstract: A representation of a virtual language teacher assists in language learning. The virtual language teacher may appear as a “talking head” in a video that a student views to practice pronunciation of a foreign language. A system for generating a virtual language teacher receives input text. The system may generate a video showing the virtual language teacher as a talking head having a mouth that moves in synchronization with speech generated from the input text. The video of the virtual language teacher may then be presented to the student.

Abstract: Frame mapping-based cross-lingual voice transformation may transform a target speech corpus in a particular language into a transformed target speech corpus that remains recognizable, and has the voice characteristics of a target speaker that provided the target speech corpus. A formant-based frequency warping is performed on the fundamental frequencies and the linear predictive coding (LPC) spectrums of source speech waveforms in a first language to produce transformed fundamental frequencies and transformed LPC spectrums. The transformed fundamental frequencies and the transformed LPC spectrums are then used to generate warped parameter trajectories. The warped parameter trajectories are further used to transform the target speech waveforms in the second language to produce transformed target speech waveform with voice characteristics of the first language that nevertheless retain at least some voice characteristics of the target speaker.

Abstract: A representation of a speech signal is received and is decoded to identify a sequence of position-dependent phonetic tokens wherein each token comprises a phone and a position indicator that indicates the position of the phone within a syllable.

Abstract: An “Animation Synthesizer” uses trainable probabilistic models, such as Hidden Markov Models (HMM), Artificial Neural Networks (ANN), etc., to provide speech and text driven body animation synthesis. Probabilistic models are trained using synchronized motion and speech inputs (e.g., live or recorded audio/video feeds) at various speech levels, such as sentences, phrases, words, phonemes, sub-phonemes, etc., depending upon the available data, and the motion type or body part being modeled. The Animation Synthesizer then uses the trainable probabilistic model for selecting animation trajectories for one or more different body parts (e.g., face, head, hands, arms, etc.) based on an arbitrary text and/or speech input. These animation trajectories are then used to synthesize a sequence of animations for digital avatars, cartoon characters, computer generated anthropomorphic persons or creatures, actual motions for physical robots, etc.

Abstract: Hidden Markov Models HMM trajectory tiling (HTT)-based approaches may be used to synthesize speech from text. In operation, a set of Hidden Markov Models (HMMs) and a set of waveform units may be obtained from a speech corpus. The set of HMMs are further refined via minimum generation error (MGE) training to generate a refined set of HMMs. Subsequently, a speech parameter trajectory may be generated by applying the refined set of HMMs to an input text. A unit lattice of candidate waveform units may be selected from the set of waveform units based at least on the speech parameter trajectory. A normalized cross-correlation (NCC)-based search on the unit lattice may be performed to obtain a minimal concatenation cost sequence of candidate waveform units, which are concatenated into a concatenated waveform sequence that is synthesized into speech.

Abstract: Techniques for providing real-time animation for a personalized cartoon avatar are described. In one example, a process trains one or more animated models to provide a set of probabilistic motions of one or more upper body parts based on speech and motion data. The process links one or more predetermined phrases that represent emotional states to the one or more animated models. After creation of the models, the process receives real-time speech input. Next, the process identifies an emotional state to be expressed based on the one or more predetermined phrases matching in context to the real-time speech input. The process then generates an animated sequence of motions of the one or more upper body parts by applying the one or more animated models in response to the real-time speech input.

Abstract: Embodiments of an audio-to-video engine are disclosed. In operation, the audio-to-video engine generates facial movement (e.g., a virtual talking head) based on an input speech. The audio-to-video engine receives the input speech and recognizes the input speech as a source feature vector. The audio-to-video engine then determines a Maximum A Posterior (MAP) mixture sequence based on the source feature vector. The MAP mixture sequence may be a function of a refined Gaussian Mixture Model (GMM). The audio-to-video engine may then use the MAP to estimate video feature parameters. The video feature parameters are then interpreted as facial movement. The facial movement may be stored as data to a storage module and/or it may be displayed as video to a display device.

Abstract: A representation of a speech signal is received and is decoded to identify a sequence of position-dependent phonetic tokens wherein each token comprises a phone and a position indicator that indicates the position of the phone within a syllable.

Abstract: Described is a bimodal data input technology by which handwriting recognition results are combined with speech recognition results to improve overall recognition accuracy. Handwriting data and speech data corresponding to mathematical symbols are received and processed (including being recognized) into respective graphs. A fusion mechanism uses the speech graph to enhance the handwriting graph, e.g., to better distinguish between similar handwritten symbols that are often misrecognized. The graphs include nodes representing symbols, and arcs between the nodes representing probability scores. When arcs in the first and second graphs are determined to match one another, such as aligned in time and associated with corresponding symbols, the probability score in the second graph for that arc is used to adjust the matching probability score in the first graph. Normalization and smoothing may be performed to correspond the graphs to one another and to control the influence of one graph on the other.