Abstract: A maximum-kurtosis, distortionless response (MKDR) technique and an extension, the maximum-kurtosis, Wiener estimate (MKWE) technique, are provided. In one form, blind estimates of the speech source's channel response are made from the microphone data and MVDR is applied. The source direction is estimated by finding weights that maximize output kurtosis, or the fourth central statistical moment, in the frequency domain. The MKWE approach approximates the Wiener filter by using MKDR-output noise power estimates to compute a Wiener post-filter. These approaches can be extended to block-adaptive versions if the speech source is not quickly moving in space.

Abstract: Disclosed herein are systems and methods for navigating electronic texts. According to an aspect, a method may include determining text subgroups within an electronic text. The method may also include selecting a text seed within one of the text subgroups. Further, the method may include determining a similarity relationship between the text seed and one or more adjacent text subgroups that do not include the selected text seed. The method may also include associating the text seed with the one or more adjacent text subgroups based on the similarity relationship to create a text cluster.

Abstract: A method and apparatus for encoding a signal is provided herein. During operation a wideband signal that is to be encoded enters a filter bank. A highband signal and a lowband signal are output from the filter bank. Each signal is separately encoded. During the production of the highband signal, a downmixing operation is implemented after preprocessing, and prior to decimating. The downmixing operation greatly reduces system complexity. In fact, it will be observed that the highest sample rate in the prior-art implementation is 64 kHz whereas the sample rate in the system described above remains at 32 kHz or below. This represents a significant complexity saving, as do the reduced number of processing blocks.

Abstract: Provided in some embodiments is a computer implemented method that includes providing script data including script words indicative of dialogue words to be spoken, providing recorded dialogue audio data corresponding to at least a portion of the dialogue words to be spoken, wherein the recorded dialogue audio data includes timecodes associated with recorded audio dialogue words, matching at least some of the script words to corresponding recorded audio dialogue words to determine alignment points, determining that a set of unmatched script words are accurate based on the matching of at least some of the script words matched to corresponding recorded audio dialogue words, generating time-aligned script data including the script words and their corresponding timecodes and the set of unmatched script words determined to be accurate based on the matching of at least some of the script words matched to corresponding recorded audio dialogue words.

Abstract: A linear prediction coefficient of a signal represented in a frequency domain is obtained by performing linear prediction analysis in a frequency direction by using a covariance method or an autocorrelation method. After the filter strength of the obtained linear prediction coefficient is adjusted, filtering may be performed in the frequency direction on the signal by using the adjusted coefficient, whereby the temporal envelope of the signal is transformed. This reduces the occurrence of pre-echo and post-echo and improves the subjective quality of the decoded signal, without significantly increasing the bit rate in a band extension technique in the frequency domain represented by SBR.

Abstract: A processing device receives a message identifier associated with an application. and identifies a first language that an operating system running on the processing device is configured to use. Responsive to determining that a first application resource bundle associated with the application and configured for the first language is not available, the processing device selects a second application resource bundle that is associated with the application and configured for a second language, identifies a first version of a message associated with the message identifier in the second application resource bundle, sends the first version of the message to an automated translator with a request to translate the first version of the message to the first language, and receives a first translated version of the message that is in the first language. The processing device may then present or store the first translated version of the message.

Abstract: A DTX decision method includes: obtaining sub-band signal(s) according to an input signal; obtaining a variation of characteristic information of each of the sub-band signals; and performing DTX decision according to the variation of the characteristic information of each of the sub-band signals. With the invention, a complete and appreciate DTX decision result is obtained by making full use of the noise characteristic in the speech encoding/decoding bandwidth and using band-splitting and layered processing. As a result, the SID encoding/CNG decoding may closely follow the characteristic variation of the actual noise.

Abstract: A speech recognition method including the steps of receiving a speech input from a known speaker of a sequence of observations and determining the likelihood of a sequence of words arising from the sequence of observations using an acoustic model. The acoustic model has a plurality of model parameters describing probability distributions which relate a word or part thereof to an observation and has been trained using first training data and adapted using second training data to said speaker. The speech recognition method also determines the likelihood of a sequence of observations occurring in a given language using a language model and combines the likelihoods determined by the acoustic model and the language model and outputs a sequence of words identified from said speech input signal. The acoustic model is context based for the speaker, the context based information being contained in the model using a plurality of decision trees and the structure of the decision trees is based on second training data.

Abstract: The audio coding method and system of lattice vector quantization is provided in the invention. The method comprises: dividing frequency domain coefficients of an audio signal for which a modified discrete cosine transform (MDCT) has been performed into a plurality of coding sub-bands, and quantizing and coding an amplitude envelope value of each coding sub-band to obtain coded bits of amplitude envelopes; performing bit allocation on each coding sub-band, and performing normalization, quantization and coding respectively on vectors in a low bit coding sub-band with pyramid lattice vector quantization and on vectors in a high bit coding sub-band with sphere lattice vector quantization to obtain coded bits of the frequency domain coefficients; multiplexing and packing the coded bits of the amplitude envelope and the coded bits of the frequency domain coefficients of each coding sub-band, then sending them to a decoding side.

Abstract: According to one embodiment, a reading aloud support apparatus includes a reception unit, a first extraction unit, a second extraction unit, an acquisition unit, a generation unit, a presentation unit. The reception unit is configured to receive an instruction. The first extraction unit is configured to extract, as a partial document, a part of a document which corresponds to a range of words. The second extraction unit is configured to perform morphological analysis and to extract words as candidate words. The acquisition unit is configured to acquire attribute information items relates to the candidate words. The generation unit is configured to perform weighting relating to a value corresponding a distance and to determine each of candidate words to be preferentially presented to generate a presentation order. The presentation unit is configured to present the candidate words and the attribute information items in accordance with the presentation order.

Abstract: Disclosed herein are systems and methods for navigating electronic texts. According to an aspect, a method may include receiving search criteria for searching an electronic text. Further, the method may include determining text subgroups within the electronic text. The method may also include determining, for each text subgroup, a similarity relationship between the search criteria and the text subgroup. Further, the method may include presenting, for each text subgroup, a graphic representing the similarity relationship between the text subgroup and the search criteria.

Abstract: Wind and other noise is suppressed in a signal by adaptively changing characteristics of a filter. The filter characteristics are changed in response to the noise content of the signal over time using a history of noise content. Filter characteristics are changed according to a plurality of reference filters, the characteristics of which are chosen to optimally attenuate or amplify signals in a range of frequencies.

Abstract: A method is disclosed in the present disclosure for recognizing emotion by setting different weights to at least of two kinds of unknown information, such as image and audio information, based on their recognition reliability respectively. The weights are determined by the distance between test data and hyperplane and the standard deviation of training data and normalized by the mean distance between training data and hyperplane, representing the classification reliability of different information. The method recognizes the emotion according to the unidentified information having higher weights while the at least two kinds of unidentified information have different result classified by the hyperplane and correcting wrong classification result of the other unidentified information so as to raise the accuracy while emotion recognition. Meanwhile, the present disclosure also provides a learning step with a characteristic of higher learning speed through an algorithm of iteration.

Abstract: In the field of audio encoding/decoding technologies, a signal de-noising method is provided. The method includes: selecting, according to a degree of inter-frame correlation of a frame where a spectral coefficient to be adjusted resides, at least two spectral coefficients having high correlation with the spectral coefficient to be adjusted; performing weighting on the at least two selected spectral coefficients and the spectral coefficient to be adjusted to acquire a predicted value of the spectral coefficient to be adjusted; and adjusting a spectrum of a decoded signal by using the acquired predicted value, and outputting the adjusted decoded signal. A signal de-noising apparatus corresponding to the signal de-noising method and an audio decoding system using the signal de-noising apparatus are also provided.

Abstract: An apparatus configured to model an encoded signal to estimate at least one distribution of the signal, rotate the signal with respect to a lattice, for lattice quantization of the signal, dependent on the at least one distribution of the signal, and quantize the signal rotated with respect to the lattice.

Abstract: In a speech recognition device and a speech recognition method, a key phrase containing at least one key word is received. The speech recognition method comprises steps: receiving a sound source signal of a key word and generating a plurality of audio signals; transforming the audio signals into a plurality of frequency signals; receiving the frequency signals to obtain a space-frequency spectrum and an angular estimation value thereof; receiving the space-frequency spectrum to define and output at least one spatial eigenparameter, and using the angular estimation value and the frequency signals to perform spotting and evaluation and outputting a Bhattacharyya distance; and receiving the spatial eigenparameter and the Bhattacharyya distance and using corresponding thresholds to determine correctness of the key phrase. Thereby this invention robustly achieves high speech recognition rate under very low SNR conditions.

Abstract: Disclosed herein are systems and methods for navigating electronic texts. According to an aspect, a method may include determining text subgroups within an electronic text. The method may also include selecting a text seed within one of the text subgroups. Further, the method may include determining a similarity relationship between the text seed and one or more adjacent text subgroups that do not include the selected text seed. The method may also include associating the text seed with the one or more adjacent text subgroups based on the similarity relationship to create a text cluster.

Abstract: A voice activity detector (100) includes a frame divider (201) for dividing frames of an input signal into consecutive sub-frames, an energy level estimator (202) for estimating an energy level of the input signal in each of the consecutive sub-frames, a noise eliminator (203) for analyzing the estimated energy levels of sets of the sub-frames to detect and eliminate from enhancement noise sub-frames and to indicate remaining sub-frames as speech sub-frames, and an energy level enhancer (205) for enhancing the estimated energy level for each of the indicated speech sub-frames by an amount which relates to a detected change of the estimated energy level for a current speech sub-frame relative to that for neighboring speech sub-frames.

Abstract: A strained-rough-voice conversion unit (10) is included in a voice conversion device that can generate a “strained rough” voice produced in a part of a speech when speaking forcefully with excitement, nervousness, anger, or emphasis and thereby richly express vocal expression such as anger, excitement, or an animated or lively way of speaking, using voice quality change. The strained-rough-voice conversion unit (10) includes: a strained phoneme position designation unit (11) designating a phoneme to be uttered as a “strained rough” voice in a speech; and an amplitude modulation unit (14) performing modulation including periodic amplitude fluctuation on a speech waveform.

Abstract: A speech recognition apparatus includes a speech input unit that receives input speech, a phoneme recognition unit that recognizes phonemes of the input speech and generates a first phoneme sequence representing corrected speech, a matching unit that matches the first phoneme sequence with a second phoneme sequence representing original speech, and a phoneme correcting unit that corrects phonemes of the second phoneme sequence based on the matching result.