EP1258865A2 - Device for improving the intelligibility of audio signals containing speech - Google Patents

Abstract

A variable high-pass filter (VHPF) (20) can be altered in its cut-off frequency (fc), for which purpose it has a controlling input where a control signal can be applied to alter the cut-off frequency. A low-pass filter (10) feeds an audio signal for improvement to the VHPF and an input terminal (1) sets up the audio signal. At the output of the VHPF there is an amplifier stage (30) to amplify the output signal from the VHPF by a factor (g) between 1.5 and 4.

Description

The invention relates to a circuit arrangement for improvement the intelligibility of speech-containing audio signals according to the features of the preamble of claim 1.

There are various ways in which the intelligibility of audio signals can be improved. One possibility is to improve the noisy signal. Another possibility is to improve signals that have been degraded by reverberation and echoes, etc. After all, a good audio signal can be changed, so
that it can be better understood by the hearing impaired. This is achieved, for example, with hearing aids. Ultimately, it is possible to change a good audio signal so that it is easier to understand when there is strong background noise.

The aim of the present invention is speech intelligibility a relatively good audio signal with unchanged Improve volume. This means equal intelligibility at a lower volume or improved intelligibility with ambient noise.

From US 5,459,813 it is known that so-called "unvoiced sounds "(e.g. consonants) from the much stronger" voiced sounds "(eg vowels). Since the" unvoiced sounds "are important for speech intelligibility, is in proposed this publication, this z. B. by clipping or to amplify amplitude compression.

In the publication "effects of amplitud distorsion upon intellegibility of speech "by J. C. Liqulider in the journal of acoustic society of america, October 1946 is a so-called "peak clipping" known. Such a "peak clipping" without ambient noise has little influence on speech intelligibility. A "peak clipping" at -20 dB still leads to an intelligibility of about 96%. The so-called "center clipping "is much worse because here the consonants removed that are particularly important for intelligibility are. "Peak clipping" at -24 dB only needs one gain of about 14 dB to achieve the same intelligibility. From the publication Elwood Kretsinger et al "The Use of fast limiting to improve the intelligence of speech in Noise ", Speech Monographs, March 1960 it is known that Consonants are about 12 dB weaker than vowels. One strengthens the consonants relative to the vowels, therefore becomes intelligibility of speech in the audio signal increased. You replace the clipper with a fast "peak limitter" (22 msec) you can increase the intelligibility even further. At -10 dBlimitting The intelligibility increased from 56% to 84%.

From publication Ian Thomas et al. "The Intelligibility of filtered-clipped speech in noise, "The Journal of the Audio Engineering Society, June 1970 it is known that the fundamental wave of an audio signal that contains speech is little Speech intelligibility helps during the first resonance frequency is very important. Therefore the signal should be in front of the Clipping to be high pass filtered.

From publication Ian Thomas et al., "Intelligibility enhancement through spectral weigthing ", Proceedings of the 1972 It is the IEEE Conference on Speech Communication and Processing known that clipping does make speech understandable increased, but the signal quality is impaired. In this Publication is therefore proposed to signal energy to shift to the significant frequency ranges.

From US 5,479,560 it is also known the audio signal to divide into several frequency bands and those frequency bands to strengthen relatively strongly with great energy and lower the others. This is suggested because language comes from a series of phonemes consists. Phonemes consist of a variety of frequencies. These are at the resonance frequencies of the mouth and throat particularly reinforced. A frequency band with such a spectral peak is called formant. Formants are special important for recognizing phonemes and thus speech. On Approach to improve speech intelligibility is therefore the peaks (formants) of the frequency spectrum of an audio signal to strengthen and weaken the valleys in between. The basic frequency is for an adult male of speech at around 60 to 250 Hz. The first four formants are at 500 Hz, 1 500 Hz, 2 500 Hz and 3 500 Hz (cf. U.S. Patent 5,459,813.

It is known from US 4,454,609, mainly the consonants to reinforce.

Finally, US 5,553,151 describes a so-called "forward masking ". Here weak consonants are replaced by the previous ones strong vowels covered over time. This release beats a relatively fast compressor with an "attack time" of approx. 10 msec. and a "release time "from approx. 75 to 150 msec.

Problematic with the previously known systems for increasing the speech intelligibility of speech in audio signals their relatively high complexity, which means that both a high software effort to calculate the individual Algorithms and a high amount of hardware is necessary. With simpler systems, however, the audio signal is changed so that the language no longer sounds very natural. Of Furthermore, the speech signal can be disturbed in simple systems be added, that of improved intelligibility can even counteract.

The aim of the present invention is therefore a circuit arrangement to improve the speech quality of audio signals to indicate that on the one hand requires little effort and on the other hand, the language still sounds natural.

This goal is achieved by a circuit arrangement with the features of claim 1 solved.

Developments of such a circuit arrangement are the subject of subclaims.

The invention is essentially based on the audio signal amplify to a predetermined factor and in a high pass to filter, the corner frequency of the high pass being regulated that the amplitude of the audio signal after the processing route equal to or proportional to the amplitude of the audio signal is at the entrance of the processing line.

With this circuit arrangement, the fundamental wave of the speech signal, which contributes relatively little to intelligibility, but possesses the greatest energy, is weakened and the usual Signal spectrum of the audio signal increased accordingly become. In addition, the amplitude of the vowels (large amplitude, low frequency) in the consonant transition range (small Amplitude, high frequency) to be lowered to the vowel to reduce so-called "backward masking". To do this, the whole Signal raised by a factor g. This factor controls the strength of the effect of signal enhancement, being meaningful Values for the factor g are between 1.5 and 4. With the circuit arrangement according to the invention are considered to be higher frequency components raised and the low-frequency fundamental wave lowered to the same extent so that the amplitude (or energy) of the audio signal remains unchanged. For signal components with small amplitudes, i.e. consonants, can be used with the circuit arrangement the corner frequency according to the present invention of the variable high pass can be lowered. Therefore in the Control, an "offset" can be added to the input signal, which is either fixed or proportional to the peak amplitude of the audio signal on the input side.

In a development of the invention it is provided that higher frequency Signal components in the audio signal are reduced. With a low pass in front of the variable high pass, interference can occur be suppressed in the signal.

In a further development of the invention it is provided that the cut-off frequency f _{c of} the variable high-pass filter has a lower limit, since the lowest frequency for speech is approx. A range of approximately 100 to 120 Hz has proven useful for a lower corner frequency.

Figur 1

die prinzipielle Schaltungsanordnung zur Verbesserung der Sprachverständlichkeit in einem Audiosignal,

Figur 2

eine Weiterbildung der Schaltungsanordnung von Figur 1,

Figur 3

eine andere Weiterbildung der Schaltungsanordnung von Figur 1, und

Figur 4

eine andere Weiterbildung der Schaltungsanordnung von Figur 1, und

Figur 5

eine vierte Weiterbildung der erfindungsgemäßen Schaltungsanordnung.

The circuit arrangement according to the invention is explained by way of example below with reference to figures. Show it:

Figure 1

the basic circuit arrangement for improving speech intelligibility in an audio signal,

Figure 2

a development of the circuit arrangement of Figure 1,

Figure 3

another development of the circuit arrangement of Figure 1, and

Figure 4

another development of the circuit arrangement of Figure 1, and

Figure 5

a fourth development of the circuit arrangement according to the invention.

In the following figures, the same reference numerals designate unless otherwise stated, the same parts with the same meaning.

The basic structure of the circuit arrangement according to the invention is shown in FIG. The circuit arrangement has a variable high-pass filter 20 whose corner frequency f _{c can be} changed. For this purpose, the variable high-pass filter 20 has a control input 21, to which a control signal for changing the corner frequency f _c can be applied. The audio signal to be improved is preferably supplied to this variable high pass 20 via a low pass 10. For this purpose, an input terminal 1 is provided for applying the audio signal. The low pass 10 does not have to be provided, but is advantageous in order to eliminate signal interference in the audio signal. At the output of the variable high pass 20 there is an amplifier stage 30 which amplifies the output signal of the variable high pass 20 by a factor g. This factor g is adjustable and is preferably between about 1.5 and 4. A gain factor that has been set is preferably no longer changed. The entire processing line consisting of variable high-pass filter 20 and amplifier 30 as well as optional low-pass filter 10 has an output terminal 2, at which the processed audio signal can be tapped as an output signal.

According to the invention, the corner frequency f _{c of} the variable high-pass filter 20 is regulated in the following manner in order to improve the intelligibility of speech within the audio signal. If the amplitude (or also energy) of the input signal at input 1 of the circuit arrangement is greater than the amplitude (or energy) at output 2 of the transmission link, then the basic frequency f _{c is} reduced. Incidentally, increased. If the amplitudes at input 1 and output 2 are the same or proportional to a predetermined factor, there is no further change in the base frequency f _c .

In Figure 2 is a further development of the circuit arrangement of Figure 1 shown. A comparator 36 is shown in FIG downstream integrator, which is preceded by a scaling factor Ki is provided. The output terminal of the integrator 40 is connected to the control input 21 of the variable high pass 20 Connection. The comparator 36 has two input terminals 34, 35 on, at the first terminal 34 the input signal and on whose terminal 35 the output signal of the transmission link is created.

The circuit arrangement of FIG. 3 differs from the circuit arrangement of FIG. 2 in that the integrator 40 is replaced by a digital circuit arrangement 60. In the digital circuit arrangement 60, the basic frequency f _c is increased or decreased by a step d in accordance with the output signal of the comparator 36, depending on whether the output signal xc at the output of the comparator 36 is greater or less than 0.

Finally, in Figure 4 is a further development of the invention Circuit arrangement shown. Continuing education consists in that to the pending at the input 34 Input signal an offset K is added. This offset can can be chosen constant or a weighted with a factor K. Output of a peak detector 70. At the peak detector 70, the audio signal is applied on the input side.

With the circuit arrangement according to the invention according to the figures 1 to 4 it is possible to set the fundamental wave of the audio signal lower and raise the remaining signal portion. Therefor the variable high-pass filter 20 is responsible.

In the event that a consonant follows a vowel in the speech signal, the circuit arrangement works as follows: A vowel is low-frequency with a large amplitude. In contrast, a consonant is high-frequency with a small amplitude. In the circuit arrangement according to the invention, the amplification factor g is set such that an amplification of 6 dB is achieved. The corner frequency of the variable high-pass filter 20 has been adjusted to this low frequency by the low-frequency vowel. The fundamental wave is so far lowered that the output amplitude is the same as the input amplitude of the audio signal, although the gain of 6 dB has been selected. If the vowel is followed by a consonant (higher frequency!), This is immediately raised by 6 dB, since the corner frequency of the high-pass filter 20 is still set to the low frequency of the vowel. The consonant is covered as less by the vowel. Only after a few milliseconds does the cut-off frequency f _{c increase} and thus also lower the consonant, so that the amplitude of the input signal is equal to the amplitude of the output signal of the processing section.

In the case of a transition from consonant to vowel, the circuit arrangement according to the invention from FIG. 1 operates as follows. The high pass filter 20 has tuned to the frequency of the consonant. The amplitude of the input signal corresponds to the amplitude of the output signal. If a vowel follows (low frequency), the relatively high cut-off frequency f _{c of} the high-pass filter 20 attenuates the vowel during the time transition and consequently does not cover the consonant. Only after a few milliseconds, the base frequency f _c is adjusted based on the control time of the control loop so that the amplitude of the input signal corresponds to the amplitude of the output signal.

Finally, the following should be noted: With a stereo signal, either each channel can receive its own control as described above or you can use a common control. Then z. B. (see FIG. 5) at input 34 = Abs (Input_Left) + Abs (Input_Right) and at input 35 = Abs (Output_Left) + Abs (Output_Right). The audio path (high pass, low pass, gain) is calculated separately for left and right, but the high passes have the same corner frequency f _c .

Claims (14)

Circuit arrangement for improving the intelligibility of speech-containing audio signals, in which frequency and / or amplitude components of the audio signal are changed according to predetermined parameters,
characterized in that the audio signal is amplified in a processing section by a predetermined factor g and is guided in a high pass (20), a corner frequency f _{c of} the high pass (20) being controllable such that the amplitude of the audio signal (2) after the processing section is equal to or proportional to the amplitude of the audio signal before the processing line. Circuit arrangement according to claim 1,
characterized in that the factor g> = 1 is selected. Circuit arrangement according to claim 1 or 2,
characterized in that the factor g is chosen approximately between 1.5 and 4. Circuit arrangement according to one of claims 1 to 3,
characterized in that when the amplitude of the input signal is greater than the amplitude of the output signal at the output of the processing section, the cut-off frequency f _{c is} reduced and, in the opposite case, is increased. Circuit arrangement according to one of claims 1 to 4,
characterized in that the change in the cut-off frequency f _c takes place incrementally, preferably in one Hz steps. Circuit arrangement according to one of claims 1 to 5,
characterized in that the cut-off frequency f _{c can be varied} in the range from approximately 100 Hz to 1 kHz. Circuit arrangement according to one of claims 1 to 6,
characterized in that the lower cut-off frequency f _c is approximately 100 to 120 Hz. Circuit arrangement according to one of claims 1 to 7,
characterized in that a low pass (10) is connected in front of the variable high pass (20). Circuit arrangement according to claim 8,
characterized in that the low pass (10) has a corner frequency at about 6 kHz. Circuit arrangement according to one of claims 1 to 9,
characterized in that a comparator (36) is coupled to a control input (21) of the variable high-pass filter (20) for changing the cut-off frequency (f _c ), the input signal of the processing path at one input (34) and the other input (35 ) the output signal of the processing line is switched. Circuit arrangement according to claim 10,
characterized in that an integrator (40) is connected between the control input (21) of the variable high-pass filter (20) and the output of the comparator (36). Circuit arrangement according to claim 10,
characterized in that between the control input (21) of the variable high-pass filter (20) and the output of the comparator (36) a digital circuit arrangement (60) is provided for incrementing the base frequency f _c in steps (d). Circuit arrangement according to one of claims 10 to 12,
characterized in that an offset is added to the input signal at an input (34) of the comparator (36). Circuit arrangement according to one of claims 10 to 13,
characterized in that the audio signal is a stereo signal and that a first input (34) of the comparator (36) the sum of the input signals for the left and right channels and that the second input (35) of the comparator (36) the sum of the output signal for the left and right channel is supplied.

Images