Alcatel

Alcatel N.v.

International Standard Electric Corporation

Alcatel Alsthom Compagnie Generale D'electricite

Standard Elektrik Lorenz Ag

Michael Walker

Hans J&uuml;rgen Matt

Armin Mann

Hans Jurgen Matt

Peter Heitkamper

Uwe Ackermann

Michael Maurer

Michael Trompf

Hans-Jürgen Matt

Fritz Weinschenk

Holger S&ouml;hnle

Gunter Kohler

Ladislav Holanda

The invention concerns a speech recognition enhancer () and a speech recognition system comprising such speech recognition enhancer (), an audio input unit () and a speech recognizer (). The speech recognition enhancer () is arranged between the audio input unit () and the speech recognizer (). The speech recognition enhancer () has a parametrizable pre-filtering unit (), a parametrizable dynamic voice level control unit (), a parametrizable noise reduction unit () and a parametrizable voice level control unit (). The parameters of these parametrizable units () are adjusted to the characteristics of the specific audio input unit () and/or the characteristics of the specific speech recognizer () for adapting the audio input unit () to the speech recognizer ().

Speech recognition enhancer

Method for the reduction of echo and/or noise signals in TK systems for the transmission of useful acoustic signals, in which, when a silence interval is present, the distorted useful signal is modified by a time-dependent control signal a(t) or by a control signal a(k) cycled in the rhythm of a scan rate f=1/T. The control signal a(k) is varied in such manner that, during the presence of speech signals in the useful signals, the amplitude of the control signal a(k) is set to a predetermined constant value cand, when a silence interval begins, the amplitude of the control signal a(k) is reduced continuously from one sample value to the next in accordance with the recurrence formula a(k+1)=a(k).&#x3b2; with &#x3b2;&#x3c;1. After the end of the silence interval, a(k) is again set equal to c.

Exponential echo and noise reduction in silence intervals

A process for determining an estimated value for the noise level n of a background noise superimposed on an acoustic useful signal is characterized in that the estimated value n(x) for a sampled input signal x(k) is defined as a value n1(x) which is determined by means of the minimum value of the quantity of all the successive maximum values of the input signal x(k) in each case found within a short time interval ts&#x2267;1 ms; that the value n1(x) is adopted as estimated value n(x) for the current noise level n when the dynamic variations of the input signal x(k) undershoot a threshold value &#x2208;; and that otherwise the estimated value determined in the preceding step is adopted unchanged as new estimated value n(x). In this way it is possible to achieve an extremely exact determination of the current noise level with very fast adaptation times which are considerably shorter than in known processes, with the need for only a relatively small computation outlay.

Adaptive noise level estimator

A process for noise reduction during the transmission of acoustic useful signals includes the following steps of: (a) determining when a speech pause is present; (b) branching the incoming TC signal from the main signal path and utilizing a Fourier transformation to generate a frequency spectrum; (c) storing in a buffer memory ( ) the last frequency spectrum recorded during the last speech pause; (d) using an inverse Fourier transformation on the respective last recorded frequency spectrum to generate a simulated noise signal; (e) subtracting the simulated noise signal in the time domain from the current incoming TC signal. As a result, the original signal is maintained uncorrupted up to the actual noise subtraction. With a simple arrangement and less computing effort than before, the process enables an overall acoustic impression to be produced, which is as agreeable as possible to the human ear and which can be matched to individual requirements. Simple optimization to the spectral processing requirements of noise signals can be realized independently of the voice signal processing requirements.

Time-domain noise suppression

Speech level measurement is particularly significant for successful echo compensation in telecommunications systems, for noise suppression in a noisy environment, for example in military vehicles, or in speech recognition and in speech coding and decoding systems. A method is indicated which permits speech levels measurement only if features of speech are recognized and interferences and speech pauses are filtered out for the measurement. To this end, speech and pause detectors and a mean value generator are utilized, the time behavior of which is largely adapted to the perception capability of the human ear. Briefly spoken vowels thus are well detected, while nasal sounds or consonants are suppressed in the case of falling levels. A speech level measuring device is indicated which provides very accurate results in a short adaptation period.

Method and circuit arrangement for speech level measurement in a speech signal processing system

A compander circuit ( ) for echo cancellation in telecommunications apparatus with an estimator module ( ) for estimating an echo coupling and a delay of a line echo and a characteristic module ( ) with an associated multiplier ( ) for controlling the short-time output level of z (k) as a function of the short-time input level of z (k) is characterized in that at all inputs and outputs of the compander circuit ( ), a plurality of N communications channels are applied and outputted, respectively, at an N-fold clock rate as compared with a single communications channel; that all memory units in the compander circuit ( ) are designed as multiplexer-buffer-demultiplexer (MBD) units ( ); and that a modulo-N counter ( ) is provided which controls all MBD units ( ) so that in each of the MBD units ( ), always the same memory cell ( ) assigned to a particular communications channel is addressed.

Multichannel echo canceller with a compander

A process for suppressing total echoes comprising a plurality of partial echoes in telecommunications (TK) equipment such as terminals, transmission systems, or switching equipment by means of adaptive FIR (&equals;Finite Impulse Response) filters that simulate the total echo and subtract it from the echo-affected useful signal transmitted to the respective TK equipment, whereby n partial FIR filters are provided, which respectively simulate a partial echo, is characterized in that m parallel, mutually independent TK channels are provided with an echo suppression function and that for this purpose a variable sub-quantity of the partial FIR filters is assigned to a TK channel to be operated, respectively, as required depending on the number of partial echoes and the size of the occurring delay times between signal and echo. This permits effective echo suppression without significant memory space and computational effort even for a large number of m parallel, mutually independent TK channels, such as must be processed in switching equipment or gateways.

Process for echo suppression with adaptive fir filters

An equipment ( ) for the suppression of whole echoes composed of several partial echoes in telecommunications (TC) equipment, such as end devices, transmission systems or exchange equipment, with the help of adaptive FIR (&equals;Finite Impulse Response) filters, which reproduce the whole echo and extract it from the echo-laden useful signal and which has been transferred to the respective TC equipment, in which the homogeneous FIR filters or partial filters ( ), each of which are able to reproduce a whole echo or partial echo, are implemented on a single ASIC (&equals;Application Specific Integrated Circuit) ( ) for the m parallel and independent TC channels n to be served; and in which the whole echoes or partial echoes to be reproduced can be calculated in real time in the ASIC ( ) and in which a digital signal processor ( ) is designed to control the ASIC ( ) and set the filter settings, especially after the coefficient required to reproduce the echo is calculated on the ASIC ( ). This calculation is carried out in the digital signal processor ( ). This makes it possible to effectively suppress echoes in a large number of m parallel and independent TC channels, as they must be processed in exchanges and network hybrids, without requiring a large amount of memory space and computational effort.

Equipment and processes for echo suppression with adaptive fir filters

To be able to successfully compensate the echos occurring in bidirectional telecommunications systems, comprising for example 2-wire/4-wire converters and/or AD/DA converters and/or satellite transmission links, the echo coupling factor and echo delay time for a system must be determined with a very high degree of accuracy and independently of the double talk situation. For this purpose, in accordance with the invention a process is provided wherein, by analyzing a dialogue, states and time segments are determined in which the probability of an incorrect evaluation of the coupling factor between the transmission paths and the echo delay time is very low. To improve the measurement accuracy, auxiliary quantities dependent upon the speech dynamics and the characteristic curve of the speech signal are introduced.

Process for determining an echo coupling factor and the echo delay time in a bidirectional telecommunications system

In accordance with the invention, from the characteristic curve of the speech signal there are derived pulse sequences which are dependent upon the frequency but not the amplitude of the speech signal, so that in the case of a real echo, independently of the attenuation, identical pulse sequences occur in the case of the transmitted signal and, with a time shift, in the case of a received echo. The correspondence between the pulse sequences of transmitted signal and received signal is an indication that the received signal is an echo of the transmitted signal.

Process for improving the echo suppression in a telecommunications system

Through crosstalk on lines, interference from current transmission lines or line echoes, the useable signal is superimposed by different interference signals. The task is to find a procedure for reducing interference which, compared to the current state of the art, can be achieved with a smaller amount of computing input and is suited to both the reduction of quasi steady-state and non-steady interference. In this, the interference of a received signal is classified with regards to characteristics in the time range as a click, crackle, rumble or noise interference signal. The time range in which the interference occurs is marked. Depending on the interference type, interference blanking and/or an interpolation of the useable signal and/or a subtraction of the interference signal from the useable signal and/or a regeneration of the useable signal is carried out.

Procedure for reducing interference in the transmission of an electrical communication signal

Digital filters are used to compensate for echoes in a transmission system in which unwanted couplings occur between a signal value in the receiving direction and a signal value in the transmitting direction. The correct adjustment of the filter coefficients has a considerable influence on the achievable echo compensation. Disturbance signals such as background noises, nonlinear effects in A/D converters or an active local speaker can briefly change the adjustment of the filter coefficients in a way so that the digital filter does not compensate for the echoes and causes additional echoes in unfavorable cases. In accordance with the invention a set of filter coefficients (C(j)) are weighted with an envelope curve (H(j)) and an error measurement (F(j)) is determined therefrom. Depending on the magnitude of the error measurement (F(j)), the filter coefficients (ci) are used for further processing or are discarded. The pertinent arrangement comprises a first digital filter (1) for which the filter coefficients (ci) are continuously calculated in accordance with the NLMS algorithm. These filter coefficients (ci) are stored, weighted and the best values are used to adjust a second digital filter (7) whose output signal (9) is subtracted from the echo-affected signal, creating an echo-free output signal (10) and thus significantly increasing the quality of a natural and undisturbed transmission, even with short-time disturbances.

Method of adaptively adjusting the coefficients of a digital filter in

When telecommunications systems with analog and digital circuit technology are interconnected, the two-wire/four-wire transitions are achieved with hybrid circuits in which mismatches cannot be avoided. They result in signal reflections in the form of interfering echoes which can considerably reduce speech intelligibility. The known methods of echo compensation for hands-free facilities are not suitable for compensating line echoes, since the signal level and the delay times of lines can be very different. A compander (3.7) is used to suppress the effect of line echoes, whose transmission characteristic, in addition to being controlled by a coupling factor between the transmit path and the receive path, and by a control signal determined by the noise level, is also controlled in the invention with the speech level that is received from the distant subscriber. Furthermore, the amplification in the linear starting range of the compander characteristic is made dependent on the ratio or comparison between the speech level and the noise level on the transmit path. The invention can be used in conjunction with hands-free facilities in communications terminals, in addition it can be spatially assigned to the hybrid circuits, or it can find central applications in base stations of radio exchange systems or in switching centers.

Method of and circuit arrangement for improving the transmission

With hands-free communication methods for stereophonic hands-free communication to transmit room acoustics, and with multichannel hands-free communication to reproduce different speaker locations at a conference table, the acoustical couplings between loudspeakers and microphones located in the room must especially be taken into consideration, in addition to the influence of ambient noise and the control of duplex communication situations. The n.sub.2 couplings take place in n-channel transmission systems. According to the invention, for each coupling, a coupling factor (dlm) is formed from the quotient of a short-term average value of the transmitted signal (ysam) assigned to a microphone, and the short-term average value of the received signal (xsam) assigned to a loudspeaker. With the coupling factor (dlm) weighted with the short-term average value (xsam), the transmission curve of a compander (1.1, 1.2) is automatically and briefly shifted in a way, so that ambient noise and echoes are suppressed in each transmission channel, and only the speech signal is transmitted.

Hands-free communication method for a multichannel transmission system

A method of adaptive echo cancellation finds application for example in the hands-free facilities of telecommunications terminals, to compensate for disturbing echoes from the acoustic feedback between loudspeaker (1) and microphone (2). It is known to simulate these echoes (h(k)) with digital filters, and to subtract them from the echo-affected microphone signal (z(k)). The realization usually requires a signal processor with a high operating speed, a floating point logic and a large memory space. The task is to present a cost-effective method, which can be used under different acoustic conditions. The invention fulfills this task by using a known FIR filter (5), whose filter coefficients (c.sub.1 to c.sub.N) are determined according to the NLMS algorithm. The step width (.alpha.), which determines the changing speed of the filter coefficients (c.sub.1 to c.sub.N), can be made a function of one or more of various factors, including the current loudspeaker-microphone coupling (dlm), a quality criterion (Q) of the excitation, the existence of local speech signals (lokspk), the existence of briefly fluctuating local noises (lnoise), and the efficiency of the echo canceller. A special scaling of the coefficient values in conjunction with the adaptive step width control, advantageously permits the use of short FIR filters and cost-effective fixed-point computers.

Growing community of inventors

Baltmannsweiler, Germany

Michael Walker