Abstract: A data processing device uses a portion of a random access memory as an output buffer for holding a frame of PCM sample data which is being output after being processed by a processing unit within the processing device. Fine grained synchronization between a reference clock and a stream of PCM data frames is provided by transferring only a portion of selected frame of PCM sample data PCM(n+1), in response to a time difference 971. A breakpoint address is determined to delineate the portion of the selected frame that is to be transferred. A sorted list of the addresses of the discontinuities is maintained in breakpoint queue. Since the buffer is managed in a FIFO manner, a single breakpoint register is sufficient to monitor addresses as they are provided by an address register for accessing the random access memory. When a breakpoint is detected, the breakpoint queue and the breakpoint register is updated by an update task 802.

Abstract: A speech coder and a method for speech coding wherein the speech signal is represented by an excitation signal applied to a synthesis filter. The speech is partitioned into frames and subframes. A classifier identifies which of several categories the speech frame belongs to, and a different coding method is applied to represent the excitation for each category. For some categories, one or more windows are identified for the frame where all or most of the excitation signal samples are assigned by a coding scheme. Performance is enhanced by coding the important segments of the excitation more accurately. The window locations are determined from a linear prediction residual by identifying peaks of the smoothed residual energy contour. The method adjusts the frame and subframe boundaries so that each window is located entirely within a modified subframe or frame.

Abstract: An audio signal encoding apparatus is constructed by: means for separating an input audio signal into a plurality of hierarchical data on the basis of a frequency; a plurality of encoding means for respectively encoding the plurality of hierarchical data; and error correction encoding means for adding an error correction code to an output of each of the plurality of encoding means, wherein the error correction encoding means raises error correcting ability which is caused by the error correction encoding for the hierarchical data on the low band side among the plurality of hierarchical data than that for the hierarchical data on the high band side. A bit allocation amount may be determined in accordance with a scale factor and a psychological auditory sensation model. Additionally, the busyness degree of a transmission path may be detected and a predetermined band or bands of the hierarchical data to be decoded or used by a reception side may be selected in accordance with the busyness degree.

Abstract: Tonal audio signals can be modeled as a sum of sinusoids with time-varying frequencies, amplitudes, and phases. An efficient encoder and synthesizer of tonal audio signals is disclosed. The encoder determines time-varying frequencies, amplitudes, and, optionally, phases for a restricted number of dominant sinusoid components of the tonal audio signal to form a dominant sinusoid parameter sequence. These components are removed from the tonal audio signal to form a residual tonal signal. The residual tonal signal is encoded using a residual tonal signal encoder (RTSE). In one embodiment, the RTSE generates a vector quantization codebook (VQC) and residual codebook sequence (RCS). The VQC may contain time-domain residual waveforms selected from the residual tonal signal, synthetic time-domain residual waveforms with magnitude spectra related to the residual tonal signal, magnitude spectrum encoding vectors, or a combination of time-domain waveforms and magnitude spectrum encoding vectors.

Abstract: An off-site data entry system comprises a computer device located at a dictation station, the computer device comprising a means to accept voice dictation from a user. The device further comprises a converter means which functions to convert the voice dictation from analog to digital status, a means to compress the voice dictation, and a means to pass the digital representation of the voice dictation to a plurality of memory registers. The system also comprises a means to telephonically transmit the voice dictation to a transcription station, the transcription station that includes at least one human transcriptionist. The system then routes the voice dictation to an available transcriptionist who has the ability to buffer the voice dictation, wherein the digital representation of the dictation is addressed by software, which allows such to be delayed, rewound, and advanced while simultaneously receiving and storing additional real-time voice dictation.

Abstract: A sound decompressing apparatus which achieves special reproducing, also known as trick play, two examples of which are forward search reproducing and reverse search reproducing, by selecting and decompressing frames containing sound data at fixed or predetermined intervals. The apparatus may adjust the output level of reproduced sounds during special reproducing.

Abstract: A nonlinear processor (NLP) for selectively removing or reducing residual echo signals from an acoustic echo canceller associated with a telephony terminal is provided. Low level background noise and near end speech signals pass through the NLP structure substantially unaltered. Distortion, background noise above a preset threshold and echo signals including long duration echoes are replaced with a linear combination of previous noise data.

Abstract: An apparatus and procedure for performing phase detection in which one-pitch cycle of an input signal waveform is cut out on a time axis. The cut-out one pitch cycle is filled with zeroes to form 2N samples (where N is an integer, 2N is equal to or greater than the number of samples of the one-pitch cycle), and the samples are subjected to an orthogonal conversion such as fast Fourier transform, whereby a real and imaginary part are used to calculate tan−1 to obtain a basic phase information. This basic phase is subjected to linear interpolation to obtain phases of respective higher harmonics of the input signal waveform.

Abstract: A method for fast match processing, comprising two stages, a pre-processing stage and an on-line stage. The pre-processing stage comprises the steps of computing an a-priori probability of occurrence for each word from an acoustic vocabulary; deriving a penalty score for each word from said acoustic vocabulary based on each words a-priori probability of occurrence in an input text. The on-line stage operates on an input text stream, comprising the steps of, computing a path score for each word from said input text; combining the computed path score with the derived penalty score to form a combined score and testing the combined score against a threshold to determine top ranking candidate words.

Abstract: Automatic speech recognition word sequence hypotheses are generated using an interleaved forward-backward search. A forward search pass uses relatively simple models for a given block period of time. A backward search pass then goes back over the previous block period using more complex models and the recognition hypotheses generated by the forward search pass. The backward search pass employs a word dependent n-best search having a flat model state organization.

Abstract: Supervised adaptation speech is supplied to the recognizer and the recognizer generates the N-best transcriptions of the adaptation speech. These transcriptions include the one transcription known to be correct, based on a priori knowledge of the adaptation speech, and the remaining transcriptions known to be incorrect. The system applies weights to each transcription: a positive weight to the correct transcription and negative weights to the incorrect transcriptions. These weights have the effect of moving the incorrect transcriptions away from the correct one, rendering the recognition system more discriminative for the new speaker's speaking characteristics. Weights applied to the incorrect solutions are based on the respective likelihood scores generated by the recognizer. The sum of all weights (positive and negative) are a positive number. This ensures that the system will converge.

Abstract: A fast and economical method for speeding up an audio signal without changing pitch can be accomplished by eliminating unneeded information from an audio signal. First, the signal is divided into chunks (frames or subframes), on which a mathematical manipulation such as a Fourier transformation is performed to identify the amplitudes of the componenet sinusoids (sines and cosines). These absolute values of the sine and cosine amplitudes for each frequency are averaged together, and the highest value(s) represents the signature, or dominant frequency/frequencies. The dominant frequency/frequencies or signatures from one chunk are compared to those of the next, and when identical the latter unit is marked as redundant. The final step consists of discarding redundant chunks from the original data, thus providing a shortened signal for replay. The pitch will not change because the only modification to the original signal was the elimination of redundant data.

Abstract: A voicing probability determination method is provided for estimating a percentage of unvoiced and voiced energy for each harmonic within each of a plurality of bands of a speech signal spectrum. Initially, a synthetic speech spectrum is generated based on the assumption that speech is purely voiced. The original and synthetic speech spectra are then divided into plurality of bands. The synthetic and original speech spectra are compared harmonic by harmonic, and a voicing determination is made based on this comparison. In one embodiment, each harmonic of the original speech spectrum is assigned a voicing decision as either completely voiced or unvoiced by comparing the difference with an adaptive threshold. If the difference for each harmonic is less than the adaptive threshold, the corresponding harmonic is declared as voiced; otherwise the harmonic is declared as unvoiced. The voicing probability for each band is then computed based on the amount of energy in the voiced harmonics in that decision band.

Abstract: A speech synthesizer and a method of synthesizing speech are provided. The speech synthesizer includes a memory unit having an interrupt vector section, a voice list section, a control program section, and a speech data section; a voice list pointer for pointing to the address in the voice list section of the memory unit where data are to be retrieved; a start address register whose content represents the starting address of a specific segment of waveform data stored in the speech data section of the memory unit; a program counter whose output is used to gain access to specific addresses in the control program section of the memory unit; a synthesizer, coupled to the memory unit, for synthesizing the retrieved speech data from the memory unit into voice data; and an interrupt controller coupled to the synthesizer, which is capable of actuating the execution of an synthesis interrupt service routine stored in the memory unit in response to an interrupt signal generated by the synthesizer.

Abstract: A speech coding and decoding system consisting of a speech coding system and a speech decoding system, the speech coding system comprises a low pass filter using an LPC parameter, and an efficient coding processing unit for generating a coded speech signal by referring to a code book for a speech signal when coding a speech and generating a noise signal by referring to the code book for the signal filtered by the low pass filter when coding the information other than a speech, the speech decoding system comprises an efficient decoding processing unit for decoding the coded signal supplied from the speech coding system so to reproduce a speech signal, and a high pass filter using the LPC parameter for filtering a speech signal of an unvoiced sound area generated by the efficient decoding processing unit.