Abstract: Methods and systems for non-negative hidden Markov modeling of signals are described. For example, techniques disclosed herein may be applied to signals emitted by one or more sources. In some embodiments, methods and systems may enable the separation of a signal's various components. As such, the systems and methods disclosed herein may find a wide variety of applications. In audio-related fields, for example, these techniques may be useful in music recording and processing, source extraction, noise reduction, teaching, automatic transcription, electronic games, audio search and retrieval, and many other applications.

Abstract: Noise robust template matching may be performed. First features of a first signal may be computed. Based at least on a portion of the first features, second features of a second signal may be computed. A new signal may be generated based on at least another portion of the first features and on at least a portion of the second features.

Abstract: Online source separation may include receiving a sound mixture that includes first audio data from a first source and second audio data from a second source. Online source separation may further include receiving pre-computed reference data corresponding to the first source. Online source separation may also include performing online separation of the second audio data from the first audio data based on the pre-computed reference data.

Abstract: Clustering and synchronizing content may include extracting audio features for each of a plurality of files that include audio content. The plurality of files may be clustered into one or more clusters. Clustering may include clustering based on a histogram that may be generated for each file pair of the plurality of files. Within each of the clusters, the files of the cluster may be time aligned.

Abstract: A method and apparatus for canceling an echo in audio communication is disclosed. The method comprises receiving an audio signal from a network and subsequently detecting a mixture audio signal comprising a target audio signal and an echo audio signal, the echo signal corresponding to the received audio signal. The method then comprises estimating the target audio signal by determining magnitude spectrograms for the mixture and received audio signals respectively, estimating a magnitude spectrogram of the target audio signal dependent on those of the mixture and received audio signal, and generating an output audio signal that estimates the target audio signal, the output audio signal being dependent on the estimated magnitude spectrogram.

Abstract: A sound mixture may be received that includes a plurality of sources. A model may be received that includes a dictionary of spectral basis vectors for the plurality of sources. A weight may be estimated for each of the plurality of sources in the sound mixture based on the model. In some examples, such weight estimation may be performed using a source separation technique without actually separating the sources.

Abstract: A system and method are described for selecting a target sound object from a sound mixture. In embodiments, a sound mixture comprises a plurality of sound objects superimposed in time. A user can select one of these sound objects by providing reference audio data corresponding to a reference sound object. The system analyzes the audio data and the reference audio data to identify a portion of the audio data corresponding to a target sound object in the mixture that is most similar to the reference sound object. The analysis may include decomposing the reference audio data into a plurality of reference components and the sound mixture into a plurality of components guided by the reference components. The target sound object can be re-synthesized from the target components.

Abstract: Methods and apparatus for relative pitch tracking of multiple arbitrary sounds. A probabilistic method for pitch tracking may be implemented as or in a pitch tracking module. A constant-Q transform of an input signal may be decomposed to estimate one or more kernel distributions and one or more impulse distributions. Each kernel distribution represents a spectrum of a particular source, and each impulse distribution represents a relative pitch track for a particular source. The decomposition of the constant-Q transform may be performed according to shift-invariant probabilistic latent component analysis, and may include applying an expectation maximization algorithm to estimate the kernel distributions and the impulse distributions. When decomposing, a prior, e.g. a sliding-Gaussian Dirichlet prior or an entropic prior, and/or a temporal continuity constraint may be imposed on each impulse distribution.

Abstract: Our invention describes a method and a system for matching securely an unknown audio recording with known audio recordings. A plurality of known audio recordings, each known audio recording associated with an index to information uniquely identifying the known audio recording is stored on a server. An unknown audio recording cross-correlated securely with each of the plurality of known audio recordings to determine a best matching known audio recording, in which the unknown audio recording and the plurality of known audio recordings are encrypted with a public key. A best matching known audio recording is determined securely according to the cross-correlation. Next, the index of the best matching known audio recording is determined securely. Finally, the information associated with the index of the best matching known audio recording is provided securely to a user of the unknown recording.

Abstract: A method expands a bandwidth of an audio signal by determining a magnitude time-frequency representation |G(?, t) for example audio signals g(t). A set of frequency marginal probabilities PG(?|z) 221 are estimated from |G(?, t)|, and a magnitude time-frequency representation |X(?, t)| is determined from an input signal audio signal x(t). Probabilities P(z), PX(z) and PX(t|z) are determined using PG(?|z)|X(?, t)|. |?(?, t)| is reconstructed according to PzPX(z)PG(?|z)PX(t|z), and |?(?, t)| is transformed to a time domain to obtain a high-quality output audio signal ?(t) corresponding to the input audio signal x(t).

Abstract: A method and system denoises a mixed signal. A constrained non-negative matrix factorization (NMF) is applied to the mixed signal. The NMF is constrained by a denoising model, in which the denoising model includes training basis matrices of a training acoustic signal and a training noise signal, and statistics of weights of the training basis matrices. The applying produces weight of a basis matrix of the acoustic signal of the mixed signal. A product of the weights of the basis matrix of the acoustic signal and the training basis matrices of the training acoustic signal and the training noise signal is taken to reconstruct the acoustic signal. The mixed signal can be speech and noise.

Abstract: A system and method recognizes speech securely using a secure multi-party computation protocol. The system includes a client and a server. The client is configured to provide securely speech in a form of an observation sequence of symbols, and the server is configured to provide securely a multiple trained hidden Markov models (HMMs), each trained HMM including a multiple states, a state transition probability distribution and an initial state distribution, and each state including a subset of the observation symbols and an observation symbol probability distribution. The observation symbol probability distributions are modeled by mixtures of Gaussian distributions. Also included are means for determining securely, for each HMM, a likelihood the observation sequence is produced by the states of the HMM, and means for determining a particular symbol with a maximum likelihood of a particular subset of the symbols corresponding to the speech.

Abstract: A method generates envelope spectra and harmonic spectra from an input broad-band training acoustic signal. Corresponding non-negative envelope bases are trained for the envelope spectra and non-negative harmonic bases are trained for the harmonic spectra using convolutive non-negative matrix factorization. Higher-band frequencies are generated for an input lower-band acoustic signal according to the non-negative envelope bases and the non-negative harmonic bases. Then, the input lower-band acoustic signal is combined with the higher-band frequencies to produce an output broad-band acoustic signal.

Abstract: A computer-implemented method includes segmenting a plurality of video frames of a sequence of video frames into a first portion that includes a selected visual object represented in the video frame and a second portion that includes a background represented in the video frame. The selected visual object is selected by using a selection envelope.

Abstract: A method detects components of a non-stationary signal. The non-stationary signal is acquired and a non-negative matrix of the non-stationary signal is constructed. The matrix includes columns representing features of the non-stationary signal at different instances in time. The non-negative matrix is factored into characteristic profiles and temporal profiles.

Abstract: A method constructs a location model for locating and tracking sources of acoustic signals. Acoustic training signals are acquired from an acoustic training source at an unknown location in an environment with an array of microphones placed at unknown positions in the environment. From each acoustic training signal, relative acoustic features are extracted to construct a location model that is trained with the relative acoustic features.

Abstract: A method and system denoises a mixed signal. A constrained non-negative matrix factorization (NMF) is applied to the mixed signal. The NMF is constrained by a denoising model, in which the denoising model includes training basis matrices of a training acoustic signal and a training noise signal and statistics of weights of the training basis matrices. The applying produces weight of a basis matrix of the acoustic signal, of the mixed signal. A product of the weights of the basis matrix of the acoustic signal and the training basis matrices of the training acoustic signal and the training noise signal is taken to reconstruct the acoustic signal. The mixed signal can be speech and noise.

Abstract: A first party has a data vector x and a second party has a classifier defined as a set of multivariate Gaussian distributions. A secure inner dot product procedure is applied to each multivariate Gaussian distribution and the data vector x to produce a vector ai for the first party and a vector bi for the second party for each application. The secure inner dot product is then applied to each vector bi and the data vector x to produce a scalar ri for the first party and a scalar qi for the second party for each application. A summed vector of elements [(a1x+q1+r1), . . . , (aNx+qN+rN)] is formed, and an index I of the summed vector for a particular element having a maximum value is the class of the data vector x.

Abstract: Our invention describes a method and a system for matching securely an unknown audio recording with known audio recordings. A plurality of known audio recordings, each known audio recording associated with an index to information uniquely identifying the known audio recording is stored on a server. An unknown audio recording cross-correlated securely with each of the plurality of known audio recordings to determine a best matching known audio recording, in which the unknown audio recording and the plurality of known audio recordings are encrypted with a public key. A best matching known audio recording is determined securely according to the cross-correlation. Next, the index of the best matching known audio recording is determined securely. Finally, the information associated with the index of the best matching known audio recording is provided securely to a user of the unknown recording.

Abstract: A method expands a bandwidth of an audio signal by determining a magnitude time-frequency representation |G(?,t) for example audio signals g(t). A set of frequency marginal probabilities PG(?|z) 221 are estimated from |G(?,t)|, and a magnitude time-frequency representation |X(?,t)| is determined from an input signal audio signal x(t). Probabilities P(z), PX(z) and PX(t|z) are determined using PG(?|z)|X(?,t)|. |?(?,t)| is reconstructed according to PzPX(z)PG(?|z)PX(t|z), and ?(?,t)| is transformed to a time domain to obtain a high-quality output audio signal ?(t) corresponding to the input audio signal x(t).