Stmicroelectronics S.r.l.

Sgs-thomson Microelectronics S.r.l.

Viviana D'Alto

Massimo Mancuso

Rinaldo Poluzzi

Danilo Pietro Pau

Giorgio Betti

Angelo Dati

Fabrizio Airoldi

Filippo Santinello

Arcangelo Ranieri Bruna

Gianguido Rizzotto

Giuseppe Spampinato

Luca Molinari

Nunziata Ivana Guarneri

Fabio Scalise

Joseph Thomas

Daniele Sirtori

Maria G Podesta

A method to detect a vocal command, the method including: analyzing audio data received from a transducer configured to convert audio into an electric signal and analyzing the data using a first neural network. The method also includes detecting a keyword from the audio data using the first neural network on the edge device, the first neural network being trained to recognize the keyword. The method further includes activating a second neural network after the keyword is identified by the first neural network and analyzing the audio data using the second neural network, the second neural network being trained to recognize a set of vocal commands. The method to detect a vocal command may also include detecting the vocal command word using the second neural network.

Vocal command recognition

Image processing circuitry processes image frames in a sequence of image frames, for example, to identify objects of interest. The processing includes filtering motion vectors associated with a current image frame, grouping the filtered motion vectors associated with the current image frame into a set of clusters associated with the current image frame, and selectively merging clusters in the set of clusters associated with the current image frame. At least one of the filtering, the grouping and the merging may be based on one or more clusters associated with one or more previous image frames in the sequence of image frames. Motion vectors included in merged clusters associated with a previous frame may be added to filtered motion vectors before grouping the motion vectors in the current frame.

Clustering method and system, corresponding apparatus and computer program product

A method of storing a data file, particularly in the MPEG format and including a flow of different frames, comprises a protection system for the data file based on a parameter stored in the data file. Advantageously, the storage method comprises selectively protecting the frames by storing parameters that are associated with corresponding different frames whose values are selected to provide a playing quality level requested by an end user. Also, a method is provided for decoding a data file, particularly of the MPEG type and including a flow of different frames, wherein the data file is stored per above.

Data-file storage, particularly for MPEG format files

A digital photography apparatus, particularly a digital still camera, includes circuit portions for acquiring a digital image, and for obtaining a compressed image. The apparatus also includes a memory for storing the compressed image. A processor is provided for obtaining a processed image and corresponding processing parameters from the image acquired. The processor supplies as an output, in a first operative condition, the processed image to be compressed by the compression circuit portion. In a second operative condition, the processor supplies as an output the image acquired to be compressed by the compression portion and the processing parameters to be stored in the memory.

Digital photography apparatus with an image-processing unit

A method and a device for motion estimated and compensated Field Rate Up-conversion (FRU) for video applications is disclosed and claimed. The invention provides for dividing an image field to be interpolated into a plurality of image blocks, where each image block includes a respective set of image elements of the image field. In one embodiment, for each image block of a subset of image blocks, a group of neighboring image blocks is selected. A motion vector for the image block is estimated that describes the movement of the image block from a previous image field to a current image field on the basis of predictor motion vectors associated to the group of neighboring image blocks. Each image element of the image block is determined by interpolation of two corresponding image elements in the previous and current image fields related by the estimated motion vector. To estimate a motion vector, the invention provides for applying each of the predictor motion vectors to the image block to determine a respective pair of corresponding image blocks in the previous and current image fields. For each of the pairs of corresponding image blocks, an error function which is the Sum of luminance Absolute Difference (SAD) between corresponding image elements in the pair of corresponding image blocks is evaluated. For each pair of the predictor motion vectors, a degree of homogeneity is also evaluated, followed by the application of a fuzzy rule having an activation level that is proportional to the degree of homogeneity of the pair of predictor motion vectors and the error functions of the pair of predictor motion vectors. An optimum fuzzy rule having the highest activation level is selected, from which the best predictor motion vector is determined, having the smaller error function of the pair associated to the optimum fuzzy rule. In most cases, the motion vector for the image block is estimated on the basis of the best predictor motion vector.

Method for motion estimated and compensated field rate up-conversion (FRU) for video applications and device for actuating such method

The level of Gaussian noise in a memory field being scanned by rows is reduced by reconstructing each pixel by fuzzy logic processors. The processors process the values of pixels neighboring the pixel being processed and belonging to a processing window defined by the last scanned row and the row being scanned, thus minimizing the memory requirement of the filtering system to a single row. The system performs an adaptive filtering within the current field itself and does not produce &ldquo;edge-smoothing&rdquo; effects as in prior adaptive filtering systems operating on consecutive fields.

Adaptive intrafield reducing of Gaussian noise by fuzzy logic processing

A circuit for enhancing chrominance transitions in a received chrominance video signal comprises a linear-phase digital filter supplied with an input stream of discrete chromatic image elements. The filter also has a high-pass transfer characteristic in a region of the frequency spectrum corresponding to an upper limit of a transmitted chrominance signal bandwidth for enhancing high-frequency components of the received chrominance signal. The circuit additionally includes non-linear digital post-processing circuitry which accepts as input an output of the filter and the input stream of discrete chromatic image elements. The non-linear post-processing circuitry acts on the output of the filter to eliminate distortions introduced in the received chrominance signal by the filter. The post-processing circuitry then detects if the received chrominance signal contains a transition pattern corresponding to predetermined patterns. If the chrominance signal matches one of the predetermined patterns, the filter is bypassed and the output of the circuit is equal to the received chrominance signal.

Circuit for enhancing chrominance transitions in real-time video

A device for filtering video images, of the type which includes first and second circuit portions, each having first and second input terminals respectively adapted to receive digitalized luminance and chrominance components of a television signal, and an output terminal coinciding with an output terminal of the device. The first and second circuit portions further include first, second and third filters, cascade coupled to one another, and a fourth filter, respectively. The first, second and third filters incorporate a computational circuit which uses a logic of the fuzzy type to process the digitalized luminance component. The fourth filter includes a computational circuit which processes the digitalized chrominance components based upon a parameter supplied from the second filter.

Methods and apparatus for filtering images using fuzzy logic

An adaptive method for suppressing video signal echoes in television equalizers including digital filters having coefficients which are updated in an adaptive and iterative manner using a modified LMS (Least Mean Square) algorithm until the difference, or output error, between a target output signal, called the reference signal, and an outgoing signal from the equalizer is gradually reduced. The method includes the steps of:

Growing community of inventors

Milan, Italy

Viviana D'Alto