Abstract: This disclosure describes rate control techniques that can improve video encoding. In particular, the described rate control techniques exploit relationships between the number of bits encoded per frame and the number of non-zero coefficients of the video blocks after quantization. The number of number of non-zero coefficients of the video blocks after quantization is referred to as rho (?). The value of ? is generally proportional to the number of bits used in the video encoding. This disclosure utilizes a relationship between ? and a quantization parameter (QP) in order to achieve rate controlled video encoding. More specifically, this disclosure provides techniques for generating a lookup table (LUT) that maps values of ? to different QPs.

Abstract: Techniques for estimating distortion due to quantization of data are described. A histogram with multiple bins may be obtained for a set of coefficients to be quantized. Distortion due to quantization of the set of coefficients may be estimated based on the histogram and average distortions for the histogram bins. The number of coefficients in each bin may be multiplied with an average distortion for the bin to obtain a per-bin distortion. The per-bin distortions for all of the bins may be accumulated and scaled with a correction factor to obtain the estimated distortion. The techniques may be used to estimate distortions for a set of coding elements. Distortion and rate may be estimated for each coding element for each of multiple quantization steps. A set of quantization steps may be selected for the set of coding elements based on the estimated distortions and the estimated rates for the set of coding elements for different quantization steps.

Abstract: An improved system for an interactive voice recognition system (400) includes a voice prompt generator (401) for generating voice prompt in a first frequency band (501). A speech detector (406) detects presence of speech energy in a second frequency band (502). The first and second frequency bands (501, 502) are essentially conjugate frequency bands. A voice data generator (412) generates voice data based on an output of the voice prompt generator (401) and audible speech of a voice response generator (402). A control signal (422) controls the voice prompt generator (401) based on whether the speech detector (406) detects presence of speech energy in the second frequency band (502). A back end (405) of the interactive voice recognition system (400) is configured to operate on an extracted front end voice feature based on whether the speech detector (406) detects presence of speech energy in the second frequency band (502).

Abstract: The disclosure describes FGS video coding techniques that use cycle-aligned fragments (CAFs). The techniques may perform cycle-based coding of FGS video data block coefficients and syntax elements, and encapsulate cycles in fragments for transmission. The fragments may be cycle-aligned such that a start of a payload of each of the fragments substantially coincides with a start of one of the cycles. In this manner, cycles can be readily accessed via individual fragments. Some cycles may be controlled with a vector mode to scan to a predefined position within a block before moving to another block. In this manner, the number of cycles can be reduced, reducing the number of fragments and associated overhead. The CAFs may be entropy coded independently of one another so that each fragment may be readily accessed and decoded without waiting for decoding of other fragments. Independent entropy coding may permit parallel decoding and simultaneous processing of fragments.

Abstract: This disclosure describes electronic video image stabilization techniques for imaging and video devices. The techniques involve determining motion and spatial statistics for individual macroblocks of a frame, and determining a global motion vector for the frame based on the statistics of each of the macroblocks. In one embodiment, a method of performing electronic image stabilization includes performing spatial estimation on each of a plurality of macroblocks within a frame of an image to obtain spatial statistics for each of the macroblocks, performing motion estimation on each of the plurality of macroblocks to obtain motion statistics for each of the macroblocks, integrating the spatial statistics and the motion statistics of each of the macroblocks to determine a global motion vector for the frame, and offsetting the image with respect to a reference window according to the global motion vector.

Abstract: An embodiment is directed to a method for selecting a predictive macroblock partition from a plurality of candidate macroblock partitions in motion estimation and compensation in a video encoder including determining a bit rate signal for each of the candidate macroblock partitions, generating a distortion signal for each of the candidate macroblock partitions, calculating a cost for each of the candidate macroblock partitions based on respective bit rate and distortion signals to produce a plurality of costs, and determining a motion vector from the costs. The motion vector designates the predictive macroblock partition.

Abstract: This disclosure describes identifying key frames from a sequence of video frames. A first set of information generated by operating on uncompressed data is accessed. A second set of information generated by compressing the data is also accessed. The first and second sets of information are used to identify key frames from the video frames.

Abstract: The motion estimation techniques and video encoding device(s) described use a two dimensional pipeline to generate accurate motion estimation parameters for a current video block. The two dimensional pipeline uses previously calculated motion estimation parameters of relevant neighboring video blocks, including a preceding video block on a same row as the current video block, prior to the generation of the accurate motion vectors, motion vector predictors, and mode decision of the current video block The use of the two dimensional pipeline allows accurate motion vector prediction from neighboring video blocks previously not available, in the computation of motion vectors. Three engines may be used in the two dimensional pipeline, a fetch engine, an integer search engine and a fractional and spatial search engine. While the fetch engine and fraction and spatial search engine operate on one row, the integer search engine operates on another row.

Abstract: The disclosure is directed to scalable motion estimation techniques for video encoding. According to the motion estimation techniques, a motion vector search is scaled according to the computing resources available. For example, the extent of the search may be dynamically adjusted according to available computing resources. A more extensive search may be performed when computing resources permit. When computing resources are scarce, the search may be more limited. In this manner, the scalable motion estimation technique balances video quality, computing overhead and power consumption. The scalable motion estimation technique may search a series of concentric regions, starting at a central anchor point and moving outward across several concentric regions. The number of concentric regions searched for a particular video frame or macroblock is adjusted according to computing resources.

Abstract: This disclosure describes rate control techniques that can improve video coding based on a “two-pass” approach. The first pass codes a video sequence using a first set of quantization parameters (QPs) for the purpose of estimating rate-distortion characteristics of the video sequence based on the statistic of the first pass. A second set of QPs can then be defined for a second coding pass. The estimated rate-distortion characteristics of the first pass are used to select Qps for the second pass in a manner that minimizes quality fluctuation between the frames of the video sequence.

Abstract: This disclosure describes rate control techniques that can improve video coding based on a “two-pass” approach. The first pass codes a video sequence using a first set of quantization parameters (QPs) for the purpose of estimating rate-distortion characteristics of the video sequence based on the statistics of the first pass. A second set of QPs can then be defined for a second coding pass. The estimated rate-distortion characteristics of the first pass are used to select QPs for the second pass in a manner that minimizes distortion of the frames of the video sequence.

Abstract: This disclosure describes video encoding techniques and video encoding devices that implement such techniques. In one embodiment, this disclosure describes a video encoding device comprising a motion estimator that computes a motion vector predictor based on motion vectors previously calculated for video blocks in proximity to a current video block to be encoded, and uses the motion vector predictor in searching for a prediction video block used to encode the current video block, and a motion compensator that generates a difference block indicative of differences between the current video block to be encoded and the prediction video block.

Abstract: This disclosure describes rate control techniques that can improve video encoding. In particular, the described rate control techniques exploit relationships between the number of bits encoded per frame and the number of non-zero coefficients of the video blocks after quantization. The number of number of non-zero coefficients of the video blocks after quantization is referred to as rho (?). The value of ? is generally proportional to the number of bits used in the video encoding. This disclosure utilizes a relationship between ? and a quantization parameter (QP) in order to achieve rate controlled video encoding. More specifically, this disclosure provides techniques for generating a lookup table (LUT) that maps values of ? to different QPs.

Abstract: Motion estimation in video compressions systems. A programmable motion estimator may be used to estimate a motion vector for a macroblock in a current frame by searching for a matching macroblock in a previous frame. A controller may be used to program the motion estimator to perform a particular search.

Abstract: A method for processing digitized speech signals by analyzing redundant features to provide more robust voice recognition. A primary transformation is applied to a source speech signal to extract primary features therefrom. Each of at least one secondary transformation is applied to the source speech signal or extracted primary features to yield at least one set of secondary features statistically dependant on the primary features. At least one predetermined function is then applied to combine the primary features with the secondary features. A recognition answer is generated by pattern matching this combination against predetermined voice recognition templates.

Abstract: A method and system for speech recognition combines different types of engines in order to recognize user-defined digits and control words, predefined digits and control words, and nametags. Speaker-independent engines are combined with speaker-dependent engines. A Hidden Markov Model (HMM) engine is combined with Dynamic Time Warping (DTW) engines.

Abstract: A method for processing digitized speech signals by analyzing redundant features to provide more robust voice recognition. A primary transformation is applied to a source speech signal to extract primary features therefrom. Each of at least one secondary transformation is applied to the source speech signal or extracted primary features to yield at least one set of secondary features statistically dependant on the primary features. At least one predetermined function is then applied to combine the primary features with the secondary features. A recognition answer is generated by pattern matching this combination against predetermined voice recognition templates.

Abstract: An improved system for an interactive voice recognition system (400) includes a voice prompt generator (401) for generating voice prompt in a first frequency band (501). A speech detector (406) detects presence of speech energy in a second frequency band (502). The first and second frequency bands (501, 502) are essentially conjugate frequency bands. A voice data generator (412) generates voice data based on an output of the voice prompt generator (401) and audible speech of a voice response generator (402). A control signal (422) controls the voice prompt generator (401) based on whether the speech detector (406) detects presence of speech energy in the second frequency band (502). A back end (405) of the interactive voice recognition system (400) is configured to operate on an extracted front end voice feature based on whether the speech detector (406) detects presence of speech energy in the second frequency band (502).

Abstract: An improved echo cancellation system (400) includes a double talk detector (406) configured for detecting a double talk condition by monitoring voice energy in a first frequency band (503). An adaptive filter (420) is configured for producing an echo signal based on a set of coefficients, and holds the set of coefficients constant when the double talk detector (406) detects the double talk condition. A microphone system (402) inputs audible signals (404) in a second frequency band (520) that is wider and overlaps the first frequency band (503). The echo signal is used to cancel echo in the input signal. A loud speaker (401) is configured for playing voice data in a third frequency band (501) essentially equal to a difference of the first and second frequency bands (502 and 503). The first and third frequency bands (503 and 501) essentially makeup the second frequency band (502).

Abstract: A voice recognition system applies user inputs to adapt speaker-dependent voice recognition templates using implicit user confirmation during a transaction. In one embodiment, the user confirms the vocabulary word to complete at transaction, such as entry of a password, and in response a template database is updated. User utterances are used to generate test templates that are compared to the template database. Scores are generated for each test template and a winner selected. The template database includes one set of speaker independent templates and two sets of speaker dependent templates.