EP1044549A1

EP1044549A1 - Telephone with means for enhancing the subjective signal impression in the presence of noise

Info

Publication number: EP1044549A1
Application number: EP99952451A
Authority: EP
Inventors: Ahmed N. Charkani El Hassani; Daniela Parayre-Mitzova
Original assignee: Cellon France SAS; Koninklijke Philips Electronics NV
Current assignee: Cellon France SAS
Priority date: 1998-09-29
Filing date: 1999-09-14
Publication date: 2000-10-18
Also published as: KR20010032522A; JP2002526983A; FR2783991A1; WO2000019686A1; TW444488B; CN1289500A

Abstract

The invention comprises the dynamic compression of an audio signal for enhancing the subjective impression of this signal in the presence of noise: the compression of the dynamic of a signal in fact corresponds to a multiplication of each sample of this signal by a gain that depends on the amplitude of said sample. The proposed compression method is completely adaptive, because it comprises a step of selecting a compression law from various possible laws as a function of the measured noise. This selection step takes into account the level of the local noise (Ni) and of the remote noise (Nr), that is, the noise contained in the received audio signal. Application: Notably mobile telephony.

Description

Telephone with means for enhancing the subjective signal impression in the presence of noise.

DESCRIPTION

FIELD OF THE INVENTION

The invention relates to an audio restoration device comprising means for measuring noise and means for compressing the dynamic of an audio signal according to a compression law selected from various possible laws. The invention also relates to an audio restoration method comprising a step for measuring noise and a step for compressing the dynamic of an audio signal according to a compression law selected from various possible laws. The invention finally relates to a telephone including such a device, or implementing such a method.

The invention finds important applications, notably for mobile telephones that are used in particularly noisy environments. When the ambient sound level becomes too high, the audio signal is immersed in the noise that renders the use of the telephone highly uncomfortable.

BACKGROUND OF THE INVENTION European patent application EP 0 661 858 A2 describes an audio restoration device that comprises means for modifying the dynamic of the received audio signal (that is to say, the ratio between the highest amplitude and the lowest amplitude of the signal) as a function of the ambient background noise.

This restoration device brings good results when the received audio signal doesn't contain too much noise, that is to say when the noise included into the received audio signal doesn't have too high amplitudes.

SUMMARY OF THE INVENTION

A first object of the invention is to propose a device which brings better results when the audio signal contains noise. This is achieve with the audio restoration devices as claimed in claim 1 of the present application. Another object of the invention is to propose an particularly efficient way of adapting the compression law as a function of the measured noise, notably the remote noise. This object is achieved with the audio restoration devices as claimed in claim 2 of the present application. Advantageously the compression ratio, the reference level and the transition threshold are functions of the measured noise, notably the remote noise.

A further object of the invention is to propose a type of compression law which is particularly efficient when the remote noise is high. To achieve this object an expansion phase is added for expanding the dynamic of the audio signal for the amplitudes that are lower than an expansion threshold (this expansion threshold is lower than the transition threshold) so as to reduce the remote noise. In an advantageous embodiment the value of this expansion threshold is a function of the measured noise.

BRIEF DESCRIPTION OF THE DRAWINGS

These and other aspects of the invention are apparent from and will be elucidated, by way of non-limitative example, with reference to the embodiment(s) described hereinafter.

In the drawings: - Fig. 1 represents an example of a telephone including an audio restoration device,

- Figs. 2, 3, 6 and 7 give examples of various types of compression laws,

- Fig. 4 is a block diagram summarizing the various steps used for the selection of a compression law, thereafter for the calculation of the gain to be applied to the audio signal in accordance with the selected compression law, and

- Fig. 5 is a diagram summarizing the steps of an audio restoration method according to the invention.

DESCRIPTION OF PREFERRED EMBODIMENT In Fig. 1 is represented an example of a telephone 1 including an audio restoration device 2. This telephone notably includes a microphone 10 connected to an analog/digital converter 20 itself connected to a speech coder 30. This speech coder 30 is connected, on the one hand, to a channel coder 40 and, on the other hand, to means 50 for measuring background noise N_\. The output of the channel coder 40 is connected to a conventional radio transceiver circuit 60. This radio transceiver circuit 60 is also connected to a channel decoder 70 that processes the signals received by the telephone. This channel decoder 70 is connected to a speech decoder 80 that produces an audio signal Uin. This speech decoder 80 is connected, on the one hand, to means 90 for measuring remote noise N_r contained in the audio signal Uin and, on the other hand, to means 100 for compressing the dynamic of the audio signal Uin. The measurements of local noise Ni and remote nose N_r performed by the noise measuring means 50 and 90 are applied to the input of the compression means 100. These measurements are used by the compression means 100 for determining the compression law to be applied to the audio signal Uin. The compression means 100 deliver an audio signal Uout that is applied to the input of a digital/analog converter 110, itself connected to an earphone 120.

Said noise measuring means comprise: - conventional means for distinguishing an only-noise signal from a signal containing speech and noise (they may be, for example, speech detection means), - means for measuring the power of the only-noise signals.

As the noise can be considered to be stationary over a period of the order of 2s (whereas speech is only stationary over a period of the order of 20ms), it is sufficient to renew the noise measurement with each only-noise signal received.

The compression means 100 have for their object to compress the dynamic of c the audio signal as a function of the local noise and, in a preferred embodiment, as a function of the measured remote noise. The compression of the dynamic of a signal in fact corresponds to a multiplication of each sample of this signal by a gain that depends on the amplitude of said sample.

In Fig. 2 are represented three compression laws of a first family of laws. This family of laws corresponds to a first type of evolution of the gain as a function of the amplitude.

In the following of the description, the references of the type X_JB are used for designating the value in dB of a variable X, whereas the references of the type X (without an index) are used for designating the linear value of a variable X. In other words, X_JB = log(X). In Fig. 2 the gain GdB is a decreasing linear function of the amplitude Uin _B-

The three gain evolution laws that have been represented are thus straight lines D; which are featured by their slope b; and by the fact that they associate all a zero gain at an amplitude level C_dB which will be called reference level in the following of the description. The equation of these straight lines D; is thus written as: D_i :G_dB = b_i .[C_dB - Uin_dB] D_; :log(G) = b_i . [log(C) - log(Uin)]= log(C/Uin)^(b'⁾

<= D_i :G = (C/Uin)^(b'⁾ (1) where G, Uin and C are the linear values of the gain G _B, of the amplitude A B and of the reference level C _B, and b; is the absolute value of the slope of the straight line D;.

Let Uinl and Uin2 be two amplitudes of the audio signal taken to the input of the compression means, and Uoutl and Uout2 the two corresponding amplitudes obtained on the output of the compression means. From the equation (1) may be derived that the following relation links the two amplitudes Uoutl and Uout2:

From this equation (2) there may be derived that any variation in the amplitude of the input signal is transferred to the output signal with a reduction factor (1-b,). This reduction factor is called compression rate and denoted τ_t (τ_t = 1-bj). The equation (1) is thus also written as:

G = (C/Uin)^ (1)

Finally, the compression effect of the dynamic of the audio signal is all the more important as the slope bj of the straight line Dj is considerable, and thus as the compression rate τ, is low. In Fig. 2 we have bi < b < b₃ and X_\ > τ₂ > τ₃.

In Fig. 3 are represented three other compression laws of a second family of laws. This second family of laws corresponds to a second type of evolution of the gain plotted against amplitude. These laws are identical to those of Fig. 2 for the amplitudes Uin_dB that are higher than a transition threshold T2_dB that is lower than or equal to C_dB- Below the transition threshold T2_dB, the gain GdB has a constant value Gmax_dβ whatever the amplitude UindB considered. In other words, in this example a limitation of the maximum gain applicable to the samples of the audio signal has been introduced. This embodiment permits the limitation of the amplification of the remote noise contained in the audio signal. Indeed, the remote noise that is present in the audio signal generally corresponds to amplitudes that are lower than the transition threshold T2. When amplifying the low amplitudes of the audio signal, one thus runs the risk of amplifying the remote noise. And this risk of amplifying the remote noise exists all the more as the compression is large, that is to say, that the compression rate is low. The value at which a choice is made to limit the maximum gain applicable to the audio signal is thus all the lower as the compression is stronger (in Fig. 3 there is Gmax₃dB > Gmax₂ B > Gmaxidβ)-

In the following will be described in detail an embodiment of the invention for a family of laws of the second type (Fig. 3). According to the invention, the parameters featuring the evolution of the gain as a function of amplitude (τ, C and T2 in the example that has just been described) are continuous or discontinuous functions of the local noise and possibly of the remote noise contained in the signal itself: T2 = f₁(N_r,N₁) C = f₂(N_r,N,) τ = f₃(N_r,N,).

As this has already been explained, the use of a low compression rate when the remote noise N_r is considerable results in an amplification of this remote noise. In this case, it is thus to be preferred to stop or reduce the compression. For this purpose, at least one of the following measures is to be taken: increase the transition threshold T2, increase the compression rate τ, diminish the reference amplitude C. These measures may be summarized by the following equatu ions

3C 3T2 do ÷^ < 0; ÷ > 0; — - > 0 (3)

3N_r 3N_r 3N_r

On the other hand, the compression is very effective when the local noise Ni is considerable. The compression then permits to rebalance the amplitudes of the audio signal by increasing the low amplitudes and by diminishing the high amplitudes relative to a reference amplitude C. In this case the perception may also be enhanced by increasing the average level of the signal. For this purpose, at least one of the following measures is to be taken: increase the reference amplitude C, diminish the transition threshold T2 to start the compression earlier, diminish the compression rate τ. These measures may be summarized by the following equations:

^ > 0; ^ < 0; ^°- < 0 (4)

3N, ΘN, ΘN,

The following functions, which are given by way of non-limiting example, fulfil the conditions posed by the equations (3) and (4): |Uϊn

T2 = f₁ (N„N₁) = with 0 < a, < 1

N.

1 + aγ

N.

Uin

C = f₂(N_r,N,) = as- with 0 < a₂ < l and a₂ ≤a, l + a₂ ^ N,

with 0 < a₃ < l

In Fig. 4 are represented in the form of blocks the various steps of an example of a gain calculation method to be applied to a sample based on noise measurements N_r and Nj. For simplicity, the invention has been explained so far by using an evolution law of the gain in dependence on the amplitude of the audio signal (equation (1)). However, it is to be preferred to replace the amplitude of the signal by its energy E so as to avoid too fast variations of the gain. The use of the energy E permits to smooth the evolution of the signal Uout and thus to avoid distortions in the speech signal. In practice, preferably an equation of the following type is used: G(k) = [C/E(k)]^(1"τ), where E(k) is the energy of the audio signal for the k^th sample of the audio signal. E(k) is obtained, for example, by the following equation: E(k) = α^« Uin(k)^« +(l-α).E(k-l), where α is a fading factor.

In practice, the energy E is obtained by filtering the amplitude • Uin* : the z-transform of the transfer function of the filter is thus written as ot/[l-(l- ).z^"'].

In Fig. 4 are represented three blocks 200, 210 and 220 for calculating parameters T2, C and τ which characterize the compression law to be used. These blocks receive on the input the measurements N_r and Ni of the remote noise and of the local noise respectively, and derive therefrom the values of the parameters T2, C and τ by applying the functions fi, f₂ and f₃. The transition threshold T2 obtained at the output of the block 200 is supplied to a calculation block 230 that calculates the maximum MAX between this threshold T2 and the energy E(k) calculated for the sample k. This maximum MAX is a calculation parameter of the gain G to be applied to the sample k, because below the transition threshold (that is to say, when MAX= T2), this gain is equal to Gmax, and, above this transition threshold (that is to say, for MAX=E(k)), it is equal to [C/E(k)]^(1"τ). The block 240 receives on the input the value of the parameters C, τ and MAX and it derives therefrom the value G of the gain to be applied to the sample k.

It will be noted that the energy E(k) may be zero in the case (normally not very frequent) of a long period of silence. If T2=0, one thus obtains an infinite gain G (G=[C/E(k)]^(1"τ)). It is thus advantageous to systematically use a non-zero threshold T2 to do away with this risk. This means that even when one does not wish to limit the maximum value of the gain to below a certain threshold T2, it is advantageous to give a very low value to T2 to avoid having an infinite gain in the case where E(k)=0.

In Fig. 5 are summarized the various steps of an audio restoration process according to the invention. In step 300 the audio signal Uin and the measurements of noise N_r and Ni are applied to the compression means 100. The next step 310 permits to make a decision to activate or deactivate the compression. In an advantageous manner: the compression is deactivated when the remote noise N_r is high, whatever the level of the local noise , and when the remote noise N_r is low or moderate, and when the local noise Ni is low;

- the compression is activated when the remote noise N_r is low or moderate and when the local noise is high or moderate.

If the compression is deactivated, the output audio signal Uout is equal to the input audio signal Uin (arrow 311). If the compression is activated (arrow 312) the next step 320 is changed to. At step 320, the amplitude • Uin* of the audio signal is calculated. Then, in step 330, this amplitude is filtered to obtain the energy E of the audio signal (the z-transform of the transfer function of the filter is written as α/[l-(l-α).z^"']). The next step 340 is the calculation step of the gain G to be applied to the input signal. This step is described in detail with reference to Fig. 4. In step 350, the audio signal Uin is multiplied by the gain G that has been calculated, so that the output signal Uout is equal to (G.Uin).

The invention is not restricted to the embodiments that have just been described. More particularly:

- It has been considered that in the whole description the measurements of noise N| and N_r were measurements made directly on the local and remote signals received by the telephone. These measurements may also be measurements of residual noise, that is, measurements of the local noise and/or remote noise after the signals received by the telephone have passed through conventional noise reduction devices. Such an embodiment permits to diminish the constraints linked with the level of the measurements Ni and N_r in the choice of the compression law. - It is also possible to calculate the parameters that characterize a family of laws by using discontinuous functions of the local noise and of the remote noise.

- The gain evolution itself may be a discontinuous function of the amplitude or energy of the audio signal. And in that case a table is advantageously used for storing the values to be assigned to the gain as a function of the compression parameters that will have been calculated.

- The step 310, which permits to deactivate the compression in certain cases, is optional.

- And it is possible to use other types of gain evolution as a function of the amplitude or energy of the audio signal. In Fig. 6 is shown a compression law that corresponds to a third type of gain evolution as a function of the amplitude. This law is identical with those represented in Fig. 3 for the amplitudes UindB that are higher than an expansion threshold Tldβ<T2 B- Below this expansion threshold TldB, the gain G_dβ is a growing linear function of the amplitude Uin_dB- In other words, in this example of embodiment is introduced an expansion of the dynamic of the audio signal for the amplitudes lower than Tl _B- This expansion enables to reduce the remote noise present in the audio signal, and thus to enhance the listening comfort for the user. The behavior of this expansion threshold Tl is identical with that of the transition threshold T2. The value of the threshold Tl may thus be given for example, by a function of the type: lUinl Tl = f₄(N_r , N, ) = !*κ- wiιh 0 < a₄ < l and a₄ > a, l + a₄ ^- N_r

In Fig. 7 is represented a compression law that corresponds to a fourth type of gain evolution as a function of the amplitude. This law is of the same type as that of Fig. 6, but the transitions between the three zones defined by the thresholds Tl_dB and T2_dβ are progressive, so that this law is described by a curve and no longer by a succession of segments of straight lines.

The case has been described where, below the transition threshold T2, the gain evolution as a function of the amplitude is constant or growing. This evolution could also be diminishing, with a less strong diminishing than above the threshold T2.

Claims

CLAIMS:

1. An audio restoration device comprising means for measuring noise and means (100) for compressing the dynamic of an audio signal (Uin) according to a compression law selected from various possible laws, characterized in that said means for measuring noise comprise means (90) for measuring a noise signal, called remote noise (N_r), in the audio signal, and said compression law is selected as a function of the measured remote noise.

2. A device as claimed in claim 1, characterized in that the compression law is determined at least by the following parameters which are functions of the measured noise:

- a compression rate (τ) that corresponds to a decreasing evolution of the gain (G) as a function of the amplitude (Uin), at least from a threshold called transition threshold (T2),

- and a reference level (C) that is higher than or equal to said transition threshold for which the gain (G) is equal to unity.

3. A device as claimed in claim 2, characterized in that below said transition threshold, the gain to be applied to the audio signal has an evolution that is constant and/or increases as a function of the amplitude of the audio signal.

4. A device as claimed in claim 2, characterized in that said transition threshold is a function of the measured noise.

5. A telephone (1) having an audio restoration device (2) that comprises means for measuring noise and means (100) for compressing the dynamic of an audio signal (Uin) according to a compression law selected from various possible laws, characterized in that said means for measuring noise comprise means (90) for measuring a noise signal, called remote noise (N_r), in the audio signal, and said compression law is selected as a function of the measured remote noise.

6. A telephone as claimed in claim 5, characterized in that the compression law is determined at least by the following parameters which are functions of the measured noise: - a compression rate (τ) that corresponds to a decreasing evolution of the gain (G) as a function of the amplitude (Uin), at least from a threshold called transition threshold (T2),

7. A telephone as claimed in claim 6, characterized in that below said transition threshold, the gain to be applied to the audio signal has an evolution that is constant and/or increases as a function of the amplitude of the audio signal.

8. A telephone as claimed in claim 6, characterized in that said transition threshold is a function of the measured noise.

9. An audio restoration method comprising a step for measuring noise and a step for compressing the dynamic of an audio signal (Uin) according to a compression law selected from various possible laws, characterized in that said step for measuring noise comprises a step for measuring a noise signal, called remote noise (N_r), in the audio signal, and said compression law is selected as a function of the measured remote noise.

10. An audio restoration method as claimed in claim 9, characterized in that the compression law is determined at least by the following parameters which are functions of the measured noise: