Abstract: The present invention provides a method and a system for determining reliable speech recognition coefficients in noisy environment, which can increase the recognition rate in a noisy environment, such as an in-car environment. The present invention utilizes the feature that most of the energy of in-car noise is concentrated in the low frequency band. Therefore, the input speech signal is filtered to remove the signal at the frequency range in which the noise energy is concentrated. Then, the energy contour of the speech signal is calculated, so as to determine the related speech recognition coefficients. Accordingly, the influence caused by the noise can be reduced, and the recognition rate for the noisy speech can be improved.

Abstract: A method of transforming a voice application program designed for US English speakers to a voice application program for UK English speakers using a computer system is described. In one embodiment, scripts and grammars associated with the voice application program are converted from US-to-UK English. The process includes spelling normalization, lexical normalization, and pronunciation conversion (including where appropriate accounting for stress shifts). The result is necessary word pronunciations for speech recognition of UK English speaker (especially for proper nouns) as well as a script that has been conformed to use UK English spelling and lexical conventions. Additionally, the script can be annotated with pronunciations as a part of the process. Further, in one embodiment a web based interface to the conversion process is provided either standalone or as part of a voice application development environment.

Abstract: A method of creating a compressed audio output signal from a series of input audio signals is disclosed and claimed. In one embodiment, the method may include, for each of the input audio signals a) precomputing a transform corresponding to the desired compression format of the output audio signal. This may be followed by b) precomputing ancillary information relating to the compression of the transformed input audio. Next, the method may include c) mixing together the transformed input signals in the transform domain to produce an output transform domain signal. The method may then include d) algorithmically combining together the precomputed ancillary information to determine a suitable decompression strategy. Lastly, the method may include e) outputting compressed audio data comprising the output transform domain signal and the combined ancillary information.

Abstract: A method and system for providing comfort noise in the non-speech periods in speech communication. The comfort noise is generated based on whether the background noise in the speech input is stationary or non-stationary. If the background noise is non-stationary, a random component is inserted in the comfort noise using a dithering process. If the background noise is stationary, the dithering process is not used.

Abstract: A system and method are provided that reduce noise in pattern recognition signals. To do this, embodiments of the present invention utilize a prior model of dynamic aspects of clean speech together with one or both of a prior model of static aspects of clean speech, and an acoustic model that indicates the relationship between clean speech, noisy speech and noise. In one embodiment, components of a noise-reduced feature vector are produced by forming a weighted sum of predicted values from the prior model of dynamic aspects of clean speech, the prior model of static aspects of clean speech and the acoustic-environmental model.

Abstract: An adaptive noise suppression system includes an input A/D converter, an analyzer, a filter, and a output D/A converter. The analyzer includes both feed-forward and feedback signal paths that allow it to compute a filtering coefficient, which is input to the filter. In these paths, feed-forward signal are processed by a signal to noise ratio estimator, a normalized coherence estimator, and a coherence mask. Also, feedback signals are processed by a auditory mask estimator. These two signal paths are coupled together via a noise suppression filter estimator. A method according to the present invention includes active signal processing to preserve speech-like signals and suppress incoherent noise signals. After a signal is processed in the feed-forward and feedback paths, the noise suppression filter estimator then outputs a filtering coefficient signal to the filter for filtering the noise out of the speech and noise digital signal.

Abstract: A method for reducing interference in acoustic signals by using of an adaptive filter method involving spectral subtraction. The inventive method enables a significant reduction of interference in acoustic signals, especially voice signals, without causing any substantial falsification of said signals such as echo or musical tones, and significantly reduces computational requirements in comparison with other methods known per se that are similarly designed to improve signal quality.

Abstract: A system for identifying a model of an acoustic system in the presence of an external noise signal is disclosed. The system includes an acoustic actuator for generating controlled sound within the acoustic system. A sensor receives the controlled sound and the external noise signal and produces a sensed signal. A control system generates a control signal in response to an error signal. The control system includes a system model for generating an estimated response signal. The control system also generates the error signal representing the difference between the sensed signal and the estimated response signal. A masking threshold generator receives the sensed signal and the error signal and produces spectral shaping parameters. A shaped signal generator for receives the spectral shaping parameters and produces a test signal which is provided as an input to the control system.

Abstract: A method which comprises forming a first noise reduction frame (18) containing speech samples; which is windowed by a first window function. For the windowed frame, noise reduction is performed for producing a second noise reduction frame (19; 45). A speech coding frame (44) to be formed comprises noise-reduced samples of at least two successive second noise reduction frames (45, 46), partly summed with one another. On the basis of said speech coding frame (44), a set of speech coding parameters pj are determined. A lookahead part (42) of the speech coding frame is at least partly formed of a first slope (41), the first slope (10, 41) comprising a set of most recent noise-reduced samples of the second noise reduction frame, not summed with the samples of any other second noise reduction frame. The method reduces the delay caused by speech coding and noise reduction.

Abstract: A short period of background noise is recorded during a call and then played back during non-speech intervals, thereby matching as nearly as possible the spectrum and amplitude of actual background noise during the call. Segments of the recording are played back in random order to mask repetition. Recording can take place more than once during a single call or take place in more than one session. In accordance with another aspect of the invention, a small amount of white noise is added to the recorded noise to improve the randomness of the sound.

Abstract: A speech compression system with a special fixed codebook structure and a new search routine is proposed for speech coding. The system is capable of encoding a speech signal into a bitstream for subsequent decoding to generate synthesized speech. The codebook structure uses a plurality of subcodebooks. Each subcodebook is designed to fit a specific group of speech signals. A criterion value is calculated for each subcodebook to minimize an error signal in a minimization loop as part of the coding system. An external signal sets a maximum bitstream rate for delivering encoded speech into a communications system. The speech compression system comprises a full-rate codec, a half-rate codec, a quarter-rate codec and an eighth-rate codec. Each codec is selectively activated to encode and decode the speech signals at different bit rates to enhance overall quality of the synthesized speech at a limited average bit rate.

Abstract: The present invention is a device for and method of detecting voice activity by receiving a signal; computing the absolute value of the signal; squaring the absolute value; low pass filtering the squared result; computing the mean of the filtered signal; subtracting the mean from the filtered result; padding the mean subtracted result with zeros to form a value that is a power of two if the result is not already a power of two; computing a DFFT of the power of two result; normalizing the DFFT result of the last step; computing a mean of the normalization; computing a variance of the normalization; computing a power ratio of the normalization; classifying the mean, variance and power ratio as speech or non-speech based on how this feature vector compares to similarly constructed feature vectors of known speech and non-speech. The voice activity detector includes an absolute value squarer; a low pass filter; a mean subtractor; a zero padder; a DFFT; a normalizer; and a classifier.

Abstract: An apparatus and method for data processing that improves estimation of spectral parameters of speech data and reduces algorithmic delay in a data coding operation. Estimation of spectral parameters is improved by adaptively adjusting a gain function used to enhance data based on whether the data contains information speech and noise or noise only. Delay is reduced by extracting coding parameters using incompletely processed data. This data is formed by multiplying a less current portion of an input data frame with a synthesis window and a more current portion of the data frame with an inverse analysis window, and performing an overlap-add process on the data frame and a similarly processed previous data frame.

Abstract: A speech coder is operable to compress digital data representing speech using a Waveform Interpolation speech coding method. The coding method is carried out on the residual signal from a Linear Predicative Coding stage. On the basis of a series of overlapping frames of the residual signal, a series of respective spectra are found. The evolution of the spectra is filtered in a multi-stage filtering process, the filtered phase data being replaced with the original phase data at the end of each stage. This is found to result in the decoder being better able to approximate the original speech signal. This is of particular utility in relation to mobile telephony.

Abstract: A signal processing unit is disclosed for selectively routing an unfiltered input signal and a noise reduced version of the unfiltered input signal to an output port in response to a noise power estimate. Routing the unfiltered input signal to the output port when the noise power estimate is less than a noise floor threshold avoids degrading the information content of an input signal having a power level close to the noise floor. A first attenuation factor and a second attenuation factor can be applied to the unfiltered input signal. A method is disclosed for parsing a signal into a plurality of frames, selecting a maximum value for each frame, and averaging the maximum values to form a noise floor threshold.

Abstract: In a technique for canceling noise from a microphone communications path, an electrical equivalence circuit which is positioned close to and electrically matched to the microphone produces a signal free reference signal. The microphone converts speech to a voice signal. An analog multiplexer alternately switches from the microphone to the electrical equivalence circuit to produce a multiplexed analog signal composed of the voice signal from the microphone and the signal free reference signal from the electrical equivalence circuit. A wire connects the output of the analog multiplexer switch to an A/D converter. The wire picks up noise from the surrounding environment. The A/D converter converts the multiplexed signal having the noise to a plurality of voice samples taken from the voice signal portion of the multiplexed signal and a plurality of noise samples taken from the signal free reference signal portion of the multiplexed signal. A noise cancellation unit applies a noise suppression algorithm (e.g.

Abstract: An apparatus and method for indicating sources of high error conditions, including out-of-range and interference conditions, to a user of a digital voice communication system. When the vocoder (206) of a communication unit (200, 450) silences due to high error conditions, a source of the high error conditions (e.g., out-of-range or interference) is determined based on the signal strength, and a source indicator (218, 220) is generated to inform the user of the source of the high error conditions. In one embodiment, where a repeater (400) is used, sources of high error conditions are detected both inbound to and outbound from the repeater. In an event of high error conditions on the inbound signal path (404), the repeater puts out a distinctive erasure (420, 422, 424) identifying the source. The communication unit (450) detects the erasure and generates a source indicator associated with the erasure.

Abstract: An improved noise reduction algorithm is provided, as well as a voice activity detector, for use in a voice communication system. The voice activity detector allows for a reliable estimate of noise and enhancement of noise reduction. The noise reduction algorithm and voice activity detector can be implemented integrally in an encoder or applied independently to speech coding application. The voice activity detector employs line spectral frequencies and enhanced input speech which has undergone noise reduction to generate a voice activity flag. The noise reduction algorithm employs a smooth gain function determined from a smoothed noise spectral estimate and smoothed input noisy speech spectra. The gain function is smoothed both across frequency and time in an adaptive manner based on the estimate of the signal-to-noise ratio. The gain function is used for spectral amplitude enhancement to obtain a reduced noise speech signal.

Abstract: The audio source is spectrally shaped by filtering in the time domain to approximate or emulate a standardized or target microphone input channel. The background level is adjusted by adding noise to the time domain signal prior to the onset of speech to set a predetermined background noise level based on a predetermined target. The audio source is then monitored in real time and the signal-to-noise ratio is adjusted by adding noise to the time domain signal, in real time, to maintain a signal-to-noise ratio based on a predetermined target value. The normalized audio signal may be applied to both training speech and test speech. The resultant normalization minimizes the mismatch between training and testing and also improves other speech processing functions, such as speech endpoint detection.

Abstract: A method and system for reducing background noise during a telephone call. When a caller begins dialing a telephone number, the system receives an ambient noise level measured near the caller's telephone. The system then computes an inverse noise waveform, which corresponds to the received ambient noise level. The inverse noise waveform is then transmitted along the same telephone line used by the completed call. The inverse noise waveform reduces the ambient background noise from the caller's location for the duration of the telephone call.

Abstract: A system and method for source coding a signal to localize transmission errors to a set of samples is disclosed. The signal comprises a plurality of signal elements (SEs) with each SE having a plurality of components. The signal is divided into a plurality of data sets with each data set having a set of SEs. Each SE component of a data set is grouped into a plurality of divisions with each SE component having a plurality of bits. The plurality of bits of the SE components are distributed from the plurality of divisions across a generated bitstream. In one embodiment, this is used in the transmission of video signals over a potentially lossy communications channel.

Abstract: A spectrum-based speech enhancement system estimates and tracks the noise spectrum of a mixed speech and noise signal. The system frames and windows a digitized signal and applies the frames to a fast Fourier transform processor to generate discrete Fourier transformed (DFT) signals representing the speech plus noise signal. The system calculates the power spectrum of each frame. The speech enhancement system employs a leaky integrator that is responsive to identified noise-only components of the signal. The leaky integrator has an adaptive time-constant which compensates for non-stationary environmental noise. In addition, the speech enhancement system identified noise-only intervals by using a technique that monitors the Teager energy of the signal. The transition between noise-only signals and speech plus noise signals is softened by being made non-binary. Once the noise spectrum has been estimated, it is used to generate gain factors that multiply the DFT signals to produce noise-reduced DFT signals.

Abstract: A novel and improved digital FM audio processor for use in a dual-mode communication system selectively operative in either FM or code division multiple access (CDMA) modes. Analog voice or voice-band data is input to a speech encoder/decoder (CODEC) which converts the analog signal to a digital signal. The digital FM signal is read from the CODEC, filtered, compressed, up-sampled and combined with a transponded SAT signal and then modulated for RF transmission. On the receive side, the FM analog signal is received, demodulated, down-sampled, expanded, and filtered before being converted to the proper format (&mgr;-law, a-law, or linear) for the speech CODEC. The CODEC then converts the digital FM audio signal into an analog waveform for conversion to sound. By performing the FM audio processing digitally, the same digital signal processing (DSP) firmware may integrated on the same application specific integrated circuit (ASIC) which is capable of performing audio processing of both FM and CDMA audio signals.

Abstract: An audio encoder 3 divides on a time axis an input audio signal into predetermined coding units and executes coding to each of the coding units so as to output a plurality of types of audio coded parameters. A cyclic redundancy check (CRC) code calculation block 5 selects important bits relative to human hearing from the audio coded parameters of the plurality of types from the audio encoder 3, and creates a CRC check code from the important bits. A convolution encoder 6 executes a convolution coding to the CRC check code and the important bits from the CRC code calculation block.

Abstract: View space representation data is produced in real time from a world space database representing terrain features. The world space database is first preprocessed. A database is formed having one element for each spatial region corresponding to a finest selected level of detail. A multiresolution database is then formed by merging elements and a strict error metric is computed for each element at each level of detail that is independent of parameters defining the view space. The multiresolution database and associated strict error metrics are then processed in real time for real time frame representations. View parameters for a view volume comprising a view location and field of view are selected. The error metric with the view parameters is converted to a view-dependent error metric. Elements with the coarsest resolution are chosen for an initial representation. Data set first elements from the initial representation data set are selected that are at least partially within the view volume.

Abstract: A sample speech is analyzed by a speech analyzing unit to obtain sample characteristic parameters, and a coding distortion is calculated from the sample characteristic parameters in each of a plurality of coding modules. The sample characteristic parameters and the coding distortions are statistically processed by a statistical processing unit to obtain a coding module selecting rule. Thereafter, when a speech is analyzed by the speech analyzing unit to obtain characteristic parameters, an appropriate coding module is selected by a coding module selecting unit from the coding modules according to the coding module selecting rule on condition that a coding distortion for the characteristic parameters is minimized in the appropriate coding module. Thereafter, the characteristic parameters of the speech are coded in the appropriate coding module, and a coded speech is obtained. When the coded speech is decoded, a reproduced speech is obtained.

Abstract: A transmission device performs encoding by a speech encoding portion including a speech encoder and error correction encoder, and transmits a continuous signal without any further processing. A reception device receives the continuous signal and performs channel decoding and speech decoding as one unit by a speech decoding portion including a soft-decision error correction decoder and a soft-decision speech decoder. Thus, a transmission and reception system performs an accurate signal reproduction without removing the signal including a normal bit error rate by correcting the error by the speech decoder.

Abstract: A method and system of processing speech information includes segmenting the speech information based upon detection of logical speech boundaries, such as isolated words, prior to compressing and/or transmitting the speech information. In one embodiment, a continuous stream of voice data is analyzed to detect signal segments containing the characteristics of an isolated word, thereby forming frames of speech information. The frames are data compressed to form packets that are transmitted to a remote site. Preferably, the packets include error checking information. In a receive mode, incoming packets are error checked prior to packet decoding. If transmission errors are detected, repairable packets may be corrected. Non-correctable errors cause generation of notice data that are used to notify a listener of the location of lost speech information. Notice data are also generated if the duration between two arriving packets exceeds a preselected threshold.

Abstract: To accurately determine rate and voice activity in moderate-to-low signal-to-noise ratios (SNRs) to maximize voice quality, system capacity and/or battery life, parameters from a noise suppression system are used as inputs to the rate determination function. Using this method, more of the speech is extracted from the background noise and a lower number of false onsets during fluctuating noise conditions compared with conventional systems are detected. The method is beneficial for voice activity detection (VAD) as well as rate determination (RDA) and unlike other RDA/VAD implementations, is independent of the type of speech coder employed (IS-127, CDG-27, IS-96 and GSM).

Abstract: Data, such as digitally coded speech signals, is transmitted so that, for each of successive frame periods, bits are formatted into a frame sequence. The bits are coded using convolutional coding. Error check bits are generated using (a) bits formatted into the first 50% of the frame sequence and (b) bits formatted into the last 25% of the frame sequence.

Abstract: A system for encoding voice while suppressing acoustic background noise and a method for suppressing acoustic background noise in a voice encoder are described herein. The voice encoder includes a sampler that captures frames of time-domain samples of an audio signal. A voice activity detector operatively coupled to the sampler determines presence or absence of speech in the current frame. A transformer is operatively coupled to the sampler for transforming the frame of time-domain audio samples into an estimate of the power spectrum of that frame. A noise model adapter operatively associated with the transformer updates a frequency-domain noise model based on the power spectrum estimate of the current frame if the voice activity detector indicates an absence of speech in this frame. A filter computation block operatively coupled to the noise model adapter and the transform computes a spectral enhancement (noise suppression) filter based on the current power spectrum estimate and the adapted noise model.

Abstract: Apparatus is disclosed for digitally encoding an input audio signal, for storage or transmission, comprising: a pitch detector for determining at least a dominant time-domain periodicity in the input signal; a generator for generating a prediction signal based on the dominant time domain periodicity of the input signal; a first discrete frequency domain transform generator for generating a frequency domain representation of the input signal; a second discrete frequency domain transform generator for generating a frequency domain representation of the prediction signal; a subtractor to subtract at least a portion of the frequency domain representation of the prediction signal from the frequency domain representation of the input signal to generate an error signal; and a generator to generate an output signal from the error signal and parameters defining the prediction signal. A corresponding decoder is also described.

Abstract: Speech is distinguished from noise by a spectral comparison of an input signal with a stored noise estimate. Updating of the noise estimate stored in a buffer is permitted during periods when speech is absent under control of an auxiliary detector. In order to improve operation in the presence of signals with strong harmonic components, e.g., signaling tones, an LPC prediction gain is computed from the input (x(i)) and a residual (y(i)) obtained from the input following filtering by a filter having a response complementary to the frequency spectrum of the input, and if the gain exceeds a threshold, buffer updating is suppressed.

Abstract: A noise suppression apparatus of the present invention includes a voice/non-voice discriminator for discriminating a frame signal divided into frames having a predetermined length; a Fourier transform unit for converting a frame signal into a spectrum; a noise spectrum estimation unit for estimating a noise spectrum of a frame judged as a non-voice signal; an amplitude spectrum subtractor for subtracting the product of an estimated noise spectrum and a predetermined coefficient from a spectrum obtained by the transform unit; an auditory correction noise adder for adding aa auditory correction noise spectrum to a spectrum outputted from the subtractor; and an inverse Fourier transform unit for performing inverse Fourier transform to an output of the adder.

Abstract: A method of transmitting digitized block coded audio signals includes forming scale factors of the digitized audio signals. The n(k-1) differences are formed from k successively in-time scale factors for each frequency sub-band or for a group of spectral values of the audio signal. The n(k-1) differences are grouped into at least two value classes. New scale factors are selected for each of the n sub-bands or spectral value groups based on a sequence of n(k-1) value classes. Identifying information, including the control information indicating at which locations in the sequence of n(k-1) value classes the selected new scale factors are disposed, is associated with each sequence of n(k-1) value classes. The associated selected new scale factors are assigned to each sequence of the sampled signal values and to the identifying information associated with each sequence of sampled signal values.

Abstract: Instability inherent in analysis-by-synthesis speech/audio codecs and caused in particular by channel errors during transmission of highly periodic signals such as high-frequency sine waves is removed. Analysis-by-synthesis techniques involve production, in response to the speech/audio signal and at regular time intervals called frames, of (a) a set of spectral parameters for use in driving a synthesis filter in view of synthesizing the speech/audio signal, and (b) a pitch gain for constructing a past-excitation-signal component supplied to the synthesis filter.

Abstract: To overcome the problem of poor representation of the background noise, the present invention includes a noise parameter generator (40) which uses a weighted average of auto-correlation values of the input signal generated during the noise-analysis phase. The weighting function gives less weight to the auto-correlations during the first few frames (as they may contain speech) and more weight to frames towards the end of this phase. Also included, to overcome the bursty nature of comfort noise, is a comfort noise generator (50) which gradually changes the nature of the signal from speech to pseudo-random noise after the speech-burst The comfort noise generator (50) of the present invention excites the auto-regressive filter corresponding to the noise model with a weighted combination of the past excitation and pseudo-random noise.

Abstract: An improved filtering method for use in an enhancement filter in a mixed excitation linear prediction (MELP) speech coder or a postfilter in a codebook excitation linear prediction (CELP) speech coder is disclosed which includes two filters. The first filter (62) has a transfer function of ##EQU1## where P is the set of prediction coefficients, .alpha. and .beta. are scaling factors, z is the inverse of the unit delay operation used in the transform representation of the transfer functions and sig-prob is signal probability estimator value and the second filter (65) has a transfer function of 1-.mu.z.sup.-1 * sig-prob, where .mu.= a scaling factor. The sig-prob is the signal probability value based on a comparison of power of the signals in a current frames to a long term estimate of noise power in signal probability estimator (63). The sig-prob value is 1 if the power of the signals is greater than the noise power plus 30 dB and the sig-prob is zero if the power is less than noise power plus 12 dB.

Abstract: The invention concerns a voice activity detection device in which an input speech signal (x(n)) is divided in subsignals (S(s)) representing specific frequency bands and noise (N(s)) is estimated in the subsignals. On basis of the estimated noise in the subsignals, subdecision signals (SNR(s)) are generated and a voice activity decision (V.sub.ind) for the input speech signal is formed on basis of the subdecision signals. Spectrum components of the input speech signal and a noise estimate are calculated and compared. More specifically a signal-to-noise ratio is calculated for each subsignal and each signal-to-noise ratio represents a subdecision signal (SNR(s)). From the signal-to-noise ratios a value proportional to their sum is calculated and compared with a threshold value and a voice activity decision signal (V.sub.ind) for the input speech signal is formed on basis of the comparison.

Abstract: The apparatus for encoding a plurality of digital information signals, having at least a first input unit for receiving a first digital information signal and a second input unit for receiving a second digital information signal. A signal combination unit combines the first and second digital information signal to generate a first combination signal. First and second data compression units compress the first and second digital information signal so as to obtain first and second data reduced digital information signals. First, second, and third masked threshold determining units determine first, second and third masked thresholds respectively from the first, second digital information signals and the combination signal respectively. A selection unit selects one masked threshold from the first and third masked threshold so as to obtain a first selected masked threshold. Preferably, the first selected masked threshold is the smallest of the first and third masked thresholds.

Abstract: The invention relates to a method and a circuit arrangement for adjusting the level and/or dynamic range of an audio signal in a transmission system and particularly in a mobile station. According to the invention, the level of acoustic noise in the environment of a terminal (10, 12) and the level and noise level of a received signal are measured (123) and the level and/or dynamic range of the reproduced signal are adjusted (121, 122) according to the results from said measurements. The solution according to the invention helps reduce the effect of noise in the signal transmitted on the transmission channel (11) and of the acoustic noise in the environment of the terminal (12) on the intelligibility of the reproduced information.

Abstract: A multi-channel signal compressor for compressing digital sound signals in the respective channels of a multi-channel sound system. The apparatus comprises a first-stage compression system and a second-stage compression system. In the first-stage compression system, a coupling circuit performs coupling between the digital sound signals of at least two of the channels to generate coupling-processed signals, one for each of the channels. A compressor circuit receives the coupling-processed signals from the coupling circuit and frequency divides each coupling-processed signal into frequency range signals in respective frequency ranges, and compresses the frequency range signals obtained by dividing each coupling-processed signal to generate a first-stage compressed signal.

Abstract: The invention relates to a voice recognition system which is robust to echoes and other background noise. An additional input signal describing a disturbance is evaluated such that during this additional recognition there is maximum suppression of information contained in the input signal. For this purpose, comparison vectors are formed which are continuously adapted to the instantaneous interference. The voice recognition system receives speech signals superimposed by noise signals. Additionally, a first spectral analysis unit produces first spectral values combined to first spectral vectors derived from disturbed speech signals. Estimates of the noise signals are also produced. A second spectral analysis unit produces second spectral values combined to second spectral vectors from the noise signal estimates.

Abstract: The invention relates to a method of noise suppression, a mobile station and a noise suppressor for suppressing noise in a speech signal. The suppressor comprises means (20, 50) for dividing the speech signal into a first amount of subsignals (X, P), which subsignals represent certain first frequency ranges, and suppression means (30) for suppressing noise in a subsignal (X, P) based upon a determined suppression coefficient (G). The noise suppressor further comprises recombination means (60) for recombining a second amount of subsignals (X, P) into a calculation signal (S), which represents a certain second frequency range, which is wider than the first frequency ranges and determination means (200) for determining a suppression coefficient (G) for the calculation signal (S) based upon the noise contained by it.

Abstract: A voice encoder which pauses outputting codewords in accordance with the absence of voice activity. An input aural signal is divided into frames and inputted to the voice encoder. The voice encoder has a voice activity detection circuit for determining at each frame whether voice activity is absent or present, a voice encoding circuit, a background noise update judging circuit for detecting a change in the characteristics of the input aural signal, and a control circuit. If the absence of voice activity is detected, the control circuit causes the frame at that time to be encoded as a background noise frame, and then pauses the operation of the voice encoding circuit. If the presence of voice activity is detected, the operation of the voice encoding circuit is resumed.

Abstract: A digital signal processor employable for utilization for speech processing or for some other pattern recognition overcomes the weaknesses of digital signal processors given the subtraction with following amount formation that must often be implemented in these applications, an auxiliary hardware is provided that contains the feature vector that is to be compared to reference feature vectors from the dictionary in a separate memory. The calculating work is thereby implemented by a separate arithmetic unit that provides a separate difference-forming and amount-forming unit for each feature comparison. The number of clock cycles of the digital signal processor required per comparison can be dramatically reduced by the invention. A suitable addressing method thereby assures that it is always corresponding features of the individual feature vectors that can be compared to one another.

Abstract: A method for reducing noise in a speech signal by controlling suppression of a predetermined band when an input speech signal has a large pitch strength. The noise reduction method is to be used in an apparatus having a signal characteristic calculating unit, an adjustment calculating unit 32, a consonant component valve (CE) and relative noise level value calculating unit, a prefilter or Hn value calculating unit, and a spectrum correcting unit as main components. The signal characteristic calculating unit derives a pitch strength of the input speech signal. The adjustment calculating unit derives an adjustment value according to the pitch strength. The CE and NR value calculating unit derives an NR value according to the pitch strength. Then, the Hn value calculating unit derives the Hn value according to the NR value and sets a noise suppression rate of the input speech signal. The spectrum correcting unit 10 reduces the noise of the input speech signal based on the noise suppression rate.

Abstract: A method of detecting double talk in an echo canceller is provided. The method includes determining far end, near end and echo signal cancelled power, determining the presence of near end, maybe near end and far end speech, and controlling an echo suppressor and filter coefficient updates in an echo canceller.

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.

Abstract: A method is disclosed for determining estimates of the perceived noise masking level of audio signals as a function of frequency. By developing a randomness metric related to the euclidian distance between (i) actual frequency components amplitude and phase for each block of sampled values of the signal and (ii) predicted values for these components based on values in prior blocks, it is possible to form a tonality index which provides more detailed information useful in forming the noise masking function. Application of these techniques is illustrated in a coding and decoding context for audio recording or transmission. The noise spectrum is shaped based on a noise threshold and a tonality measure for each critical frequency-band (bark).