Mitsubishi Electric Research Laboratories, Inc.

Mitsubishi Electric Research Lab, Inc.

Mitsubishi Electric Corporation

Bhiksha Ramakrishnan

Paris Smaragdis

Bent Schmidt-Nielsen

Rita Singh

Bret Harsham

Kaustubh Prakash Kalgaonkar

Garrett Weinberg

Peter P Wolf

Shantanu Rane

Rahul Sarpeshkar

Ajay Divakaran

Paul H Dietz

Kevin William Wilson

Lorenzo Turicchia

Manas Ashok Pathak

Evandro Bacci Gouvêa

Joseph K Woelfel

Aarthi M Reddy

Edward W D Whittaker

Evandro B Gouvea

Bret A Harsham

Embodiments of the invention disclose a system and a method for determining a differentially private aggregate classifier for a set of databases, wherein each database in the set of databases is associated with a classifier and a noise value, wherein the classifier and the noise value are determined locally for each database, such that a combination of the classifier and the noise value ensure a differential data privacy of the database, and wherein the differentially private aggregate classifier preserves the differential data privacy of each database. The differentially private aggregate classifier is a combination of the classifiers of the set of databases modified with the noise value corresponding to a smallest database in the set of databases.

Differentially private aggregate classifier for multiple databases

A method and system recognizes an unknown speaker by directing an ultrasonic signal at a face of the unknown speaker. A Doppler signal of the ultrasonic signal is acquired after reflection by the face, and Doppler features are extracted from the reflected Doppler signal. The Doppler features are classified using Doppler models storing the Doppler features and identities of known speakers to recognize and identify the unknown speaker.

Ultrasonic doppler sensor for speaker recognition

An information retrieval system stores and retrieves documents using particles and a particle-based language model. A set of particles for a collection of documents in a particular language is constructed from training documents such that a perplexity of the particle-based language model is substantially lower than the perplexity of a word-based language model constructed from the same training documents. The documents can then be converted to document particle graphs from which particle-based keys are extracted to form an index to the documents. Users can then retrieve relevant documents using queries also in the form of particle graphs.

Method for indexing for retrieving documents using particles

Method for expanding audio signal bandwidth

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.

Denoising acoustic signals using constrained non-negative matrix factorization

A computer implemented method facilitates a user interaction via a speech-based user interface. The method acquires spoken input from a user in a form of a phrase of one or more words. It further determines, using a plurality of different domains, whether the phrase is a query or a command. If the phrase is the query the method retrieves and presents relevant items from a plurality of databases. If the phrase is a command, the method performs an operation.

Method for interacting with users of speech recognition systems

An acoustic tone at a predetermined frequency is detected after being reflected and modulated by an unknown moving object. Acoustic features are extracted from the reflected acoustic tone using acoustic spectral analysis. A dimensionality of the acoustic features is reduced, and the reduced dimensionality features are statistically classified to identify the object.

Method and system for identifying moving objects using acoustic signals

A method and system enhances an acoustic signal acquired by a microphone from an acoustic source while concurrently acquiring a Doppler signal from moving parts of the acoustic source. The acoustic signal and the Doppler signal are then analyzed according to a model to generate an enhanced acoustic signal.

System and method for acquiring acoustic signals using doppler techniques

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.

Constructing broad-band acoustic signals from lower-band acoustic signals

A method and system processes a speech signal. A fast Fourier transform is performed on a speech signal to produce a speech signal having a plurality of frequency bands in a frequency domain. For each frequency bands, filter the speech signal in the frequency domain with a spatial broadband filter, compress the broadband filtered speech signal, filter the compressed speech signal with a spatial narrowband filter; and, expand the narrowband filtered signal to an expanded signal

Method and system for FFT-based companding for automatic speech recognition

A method indexes and retrieves documents stored in a database. A document feature vector is extracted from each document and the documents are then indexed according to the feature vectors. A spoken query is converted to an intermediate representation representing likelihoods of possible sequential combinations of terms in the spoken query. A query certainty vector is generated from the intermediate representation. Other information is acquired. The other information is combined with the query certainty vector. The query vector and the other information are then compared to each of the document feature vectors to retrieve a ranked result set of documents.

Method and system for retrieving documents with spoken queries

A method according to the invention separates multiple audio signals recorded as a mixed signal via a single channel. The mixed signal is A/D converted and sampled. A sliding window is applied to the samples to obtain frames. The logarithms of the power spectra of the frames are determined. From the spectra, the a posteriori probabilities of pairs of spectra are determined. The probabilities are used to obtain Fourier spectra for each individual signal in each frame. The invention provides a minimum-mean-squared error metho or a soft mask method for making this determination. The Fourier spectra are inverted to obtain corresponding signals, which are concatenated to recover the individual signals.

Separating multiple audio signals recorded as a single mixed signal

A method and system detect speech activity. An ultrasonic signal is directed at a face of a speaker over time. A Doppler signal of the ultrasonic signal is acquired after reflection by the face. Energy in the Doppler signal is measured over time. The energy over time is compared to a predetermined threshold to detect speech activity of the speaker in a concurrently acquired audio signal.

Ultrasonic Doppler sensor for speech-based user interface

A method decomposes input data acquired of a signal. An input signal is sampled to acquire input data. The input data is represented as a probability distribution. An expectation-maximization procedure is applied iteratively to the probability distribution to determine components of the probability distributions.

Shift-invariant probabilistic latent component analysis

A method classifies data into multiple classes so that the data in each class have a class-conditional probability distribution. The class-conditional probability distributions of measured data are projected into a likelihood space. The projected class-conditional probability distributions in the likelihood space are then classified according to a discriminant classifier in likelihood space.

Growing community of inventors

Watertown, MA, United States of America

Bhiksha Ramakrishnan