Abstract: A method for detecting a transient in a discrete-time audio signal is performed completely in the time domain and includes the step of segmenting the discrete-time audio signal so as to generate consecutive segments of the same length with unfiltered discrete-time audio signals xs(T−1). The discrete-time audio signal in a current segment is subsequently filtered. Then either the energy of the filtered discrete-time audio signal in the current segment can be compared with the energy of the filtered discrete-time audio signal in a preceding segment or a current relationship between the energy of the filtered discrete-time audio signal in the current segment and the energy of the unfiltered discrete-time audio signal in the current segment can be formed and this current relationship compared with a preceding corresponding relationship. On the basis of the one and/or the other of these comparisons it is detected whether a transient is present in the discrete-time audio signal.

Abstract: Classifying a call to a called destination endpoint by a call classifier. The call classifier is responsive to information received from the called destination endpoint to perform the call classification.

Abstract: A multi-rate speech codec supports a plurality of encoding bit rate modes by adaptively selecting encoding bit rate modes to match communication channel restrictions. In higher bit rate encoding modes, an accurate representation of speech through CELP (code excited linear prediction) and other associated modeling parameters are generated for higher quality decoding and reproduction. To achieve high quality in lower bit rate encoding modes, the speech encoder departs from the strict waveform matching criteria of regular CELP coders and strives to identify significant perceptual features of the input signal. The encoder generates pluralities of codevectors from a single, normalized codevector by shifting or other rearrangement. As a result, searching speeds are enhanced, and the physical size of a codebook built from such codevectors is greatly reduced.

Abstract: A method for detecting a transient in a discrete-time audio signal is performed completely in the time domain and includes the step of segmenting the discrete-time audio signal as to generate consecutive segments of the same length with unfiltered discrete-time audio signals. The discrete-time audio signal in a current segment is filtered. Either the energy of the filtered discrete-time audio signal in the current segment is compared with the energy of the filtered discrete-time audio signal in a preceding segment or a current relationship between the energy of the filtered discrete-time audio signal in the current segment and the energy of the unfiltered discrete-time audio signal in the current segment is formed and this current relationship compared with a preceding corresponding relationship. Whether a transient is present in the discrete-time audio signal is detected using one and/or the other of these comparisons.

Abstract: Phoneme analysis is carried out in real time by detecting a voiced component in the range of 200 Hz to 1 KHz and simultaneously detecting voiceless components having frequencies greater than about 2.4 KHz and greater than about 3.4 KHz, respectively, to produce respective outputs which are logically combined to produce two-bit logic signals which can be used to control a speech processing device.

Abstract: A non-coherent frequency shift keying detection scheme. The invention has been applied to the special case of capturing caller ID information when the voice-band modem is in the low power mode. The technique provides the capability for capturing and decoding enough of the caller ID information transmitted by the telephone company while the computer is in the sleep or low power mode, and upon awakening the computer can use this captured information to decode the actual caller ID information. Additionally, this invention automatically accommodates both the Bell 202 or V.23 transmission standards by utilizing a threshold concept in which the time that has elapsed between the zero crossings of an FSK tone half cycle are measured. The elapsed times may be used as an indication of which FSK tone has been transmitted thus allowing capture of caller ID information. The bit output of the caller ID capture circuit is output for storage by RAM thus allowing use by the computer when it wakes up.

Abstract: Pitch is tracked for a selected source process characterized by a pitch source having many harmonics followed by a bandpass filtering (e.g., human speech or other common processes). The filtering in the original source process causes an original pitch pulse to be seen in somewhat modified form and followed by ringing at band pass filter frequencies. Often, the ringing produces peaks of unpredictable amplitude, a characteristic making it difficult to use simplistic methods such as picking waveform amplitude peaks. The method of the present invention avoids such difficulties by taking into account relative phase of harmonics associated with the basic pitch rate or frequency (F0). Since the bandpass filters in the original process produce ringing in frequencies other than the original fundamental frequency, the instantaneous phase of each of the ringing frequencies are only temporally aligned or lined up well for the duration of the original pitch pulse (i.e.

Abstract: A human speech detection method detects pure-speech signals in an audio signal containing a mixture of pure-speech and non-speech or mixed-speech signals. The method accurately detects the pure-speech signals by computing a novel Valley Percentage feature from the audio signal and then classifying the audio signals into pure-speech and non-speech (or mixed-speech) classifications. The Valley Percentage is a measurement of the low energy parts of the audio signal (the valley) in comparison to the high energy parts of the audio signal (the mountain). To classify the audio signal, the method performs a threshold decision on the value of the Valley Percentage. Using a binary mask, a high Valley Percentage is classified as pure-speech and a low Valley Percentage is classified as non-speech (or mixed-speech). The method further employs morphological filters to improve the accuracy of human speech detection.

Abstract: The invention relates to a device intended for detecting in successive frames containing voice signals mixed with noise from various sources the periods of speech and those of only noise. By calculating for each frame its energy and the zero-crossing rate of its centered noise signal and by comparing these magnitudes with adaptive threshold values, the real state of the device is detected, which leads to specific controls adapted for each state.

Abstract: A method to detect a normalized data field of all zeros or all ones includes receiving a control field and a data field, dividing the data field into segments, and performing detections on each segment. Each segment undergoes all zeros detection, all ones detection, modified zeros detection, and modified ones detection. The modified zeros detection and modified ones detection are both done based on the control field. Each detection for each segment generates a response. Then, a pair of the four responses, or a clear responses signal, is selected for each of the segments based on the control field. From the selected responses, the method determines if the normalized data field is all zeros or all ones.

Abstract: The system and method of the present invention uses a zero-crossing rate measurement in order to determine the initiation and/or termination of speech in an audio signal input. It is especially well suited for detecting the termination of a telephone message in a telephone answering device. Specifically, a sample of the zero-crossing rate signal is determined by counting the number of consecutive speech samples required for the occurrence of a pre-defined number of consecutive zero-crossings. The resultant zero-crossing rate signal is smoothed and applied to a differentiator. A short-time magnitude integration is performed to measure the energy in the differentiated signal. The output of the magnitude integration is provided to a threshold detector which produces a sequence of decision values indicating the presence or absence of speech. Finally, the decision values are filtered to produce a more definitive sequence of final decision values.

Abstract: A method an apparatus for modifying the play time of digital audio tracks. For each data block of samples taken from a plurality of such blocks comprising an audio stream or track, the original block data plus a time shifted copy of the block data are superimposed, in a manner dependent on either desired expansion or desired contraction of play time, to create an overlap region. The overlap region is selected at a position of best match as resulting from a normalized correlation over the overlap, the super position being weighted by a linear cross-fading function. All signal crossovers (signal transitions of opposite signal sign) in a block are located. A position at which to superimpose a copy of the data block with the data block, and the length of the data block, are determined based upon the location of said crossovers within a given data block.

Abstract: A method and a device for discriminating a voiced sound from an unvoiced sound or background noise in speech signals are disclosed. Each block or frame of input speech signals is divided into plural sub-blocks and the standard deviation, effective value or the peak value is detected in a detection unit for detecting statistical characteristics from one sub-block to another. A bias detection unit detects a bias on the time scale of the standard deviation, effective value or the peak value to decide whether the speech signals are voiced or unvoiced from one block to another.

Abstract: Musical intervals are used as codes to communicate preselected commands or inputs to a computer. The system can use human voice sounding notes (DO-RE, etc.) successively to establish the interval and can use electrical circuitry and programs for converting the received notes or tones into pulse trains of the same fundamental period and for storing and calculating the period and interval. When a particular interval is received the apparatus including a digital micro-computer such as the Apple II PLUS.TM. may execute a pre-set recorded subroutine and restore the circuitry for the reception of a second interval.

Abstract: A method and system for implementing time scale modification wherein the method includes a Zero Crossing Module (22) for determining zero crossing points in the signal, a Feature Vector Module (24) for generating feature vectors describing the zero crossing points, a Distance Metric Module (26) for generating distance metrics describing local characteristics at the zero crossing points, an Alignment Module (28) for using the feature vectors and distance metrics for aligning and synchronizing the signal in accordance with local similarities and similarity over a selected time interval to generate a time scale modified signal. The present invention also includes a Cross Fade Module (20) for smoothing transitions between successive frames of the resulting time scale modified signal.

Abstract: A method and system for automatically generating at least one new voice file corresponding to at least one new text from a script incorporating a plurality of known text having corresponding preexisting voice files associated therewith. A plurality of phonetic sequences corresponding to the plurality of known text is stored in a first memory. A text input corresponding to a textual version of the script is provided and a text-to-phonetic translator translates the text input to obtain a corresponding textual phonetic sequence based on the plurality of phonetic sequences stored in the first memory. An audio input of the script is provided and a speech recognizer generates an audio phonetic sequence of the audio input. A text-to-speech aligner aligns the text input and the corresponding textual phonetic sequence with the audio input and the corresponding audio phonetic sequence to obtain an alignment of the text input and the audio input. The at least one new voice file is generated based on the alignment.

Abstract: A speech coding system employs measurements of robust features of speech frames whose distribution are not strongly affected by noise/levels to make voicing decisions for input speech occurring in a noisy environment. Linear programing analysis of the robust features and respective weights are used to determine an optimum linear combination of these features. The input speech vectors are matched to a vocabulary of codewords in order to select the corresponding, optimally matching codeword. Adaptive vector quantization is used in which a vocabulary of words obtained in a quiet environment is updated based upon a noise estimate of a noisy environment in which the input speech occurs, and the “noisy” vocabulary is then searched for the best match with an input speech vector. The corresponding clean codeword index is then selected for transmission and for synthesis at the receiver end.