Methods for processing an audio signal captured by a microphone

Abstract

A method for audio post-processing of an audio signal acquired by a microphone (200) is provided. The method comprises the following steps: determining the reverberation properties of a room (100) by acquiring a first audio signal as a test signal by at least one microphone (200) in the room and determining the reverberation properties of the room (100) based on the acquired first audio signals of the at least one microphone (200); acquiring a second audio signal by a microphone (200) in the room, wherein the second audio signal includes a desired sound and room reverberation generated by the desired sound in the room; creating a prediction of room reverberation based on the determined reverberation properties of the room and the acquired second audio signal; analyzing the second audio signal for points in the audio signal where the level of the room reverberation is higher than a level of a desired signal in the second audio signal; and outputting the second audio signal.If the level of the room reverberation is lower than the level of the desired signal, the second audio signal is output at a reduced volume if the level of the room reverberation is higher than the level of the desired signal.

Description

[0001] The present invention relates to a method for audio processing of an audio signal captured by a microphone.

[0002] Microphones are used to capture an audio signal, especially a user's speech. However, the microphone records both the user's speech and ambient noise. However, this ambient noise is largely undesirable in the captured audio signal.

[0003] It is therefore an object of the present invention to provide a method for audio processing of an audio signal detected by a microphone, which method enables improved detection of a useful signal (e.g. a speech signal) in an environment in which ambient noise is present.

[0004] This object is achieved by a method for audio processing of an audio signal according to claim 1.

[0005] Thus, a method is provided for audio post-processing of an audio signal that has been recorded by a microphone. Here, the reverberation properties of a room can first be determined. This is done by recording a first audio signal as a test signal using at least one microphone located in the room. The reverberation properties of the room are determined based on the recorded first audio signals from the at least one microphone. A second audio signal can be recorded by the microphone in the room. The second audio signal has a wanted sound and a room reverberation generated in the room by the wanted sound. A prediction of a room reverberation is created based on the determined reverberation properties of the room and the recorded second audio signal. The second audio signal is analyzed for points at which the room reverberation has a higher level than a wanted signal, in particular a speech signal.The volume of a second audio signal to be output is reduced at those points where the prediction of the room reverberation indicates that the room reverberation exceeds a threshold. In other words, the volume of the second audio signal to be output is reduced as a function of the room reverberation. In particular, if the useful signal in the second audio signal has a low level while the room reverberation has a high level, the volume of the second audio signal to be output can be reduced in order to reduce the influence of the room reverberation on the entire second audio signal to be output. The volume or level of the second audio signal to be output is thus reduced as a function of the room reverberation. At points where the useful signal (e.g. the speech signal) has a higher level than the room reverberation, the volume does not need to be reduced because the level of the useful signal exceeds the level of the room reverberation.

[0006] The wanted signal can optionally be a speech signal or an audio signal generated by a user (e.g., singing). The wanted signal represents the desired portion of the captured audio signal.

[0007] According to one aspect of the present invention, a measurement of linear properties of the system consisting of the room and the microphone can be performed. This can be done, for example, using an impulse response. Based on these measurements, a reverberation time of the room in which the microphone is located can be determined.

[0008] According to one aspect of the present invention, an algorithm is provided which follows the envelope of the input signal and detects transients and their volume or level.

[0009] According to one aspect of the present invention, the volume of the second audio signal to be output is reduced if, based on the algorithm, it has been detected that room reverberation and only a small portion of the useful signal are present in the audio signal. In particular, the volume reduction can occur if the signal level of the reverberation is below a threshold value compared to the transient level.

[0010] According to one aspect of the present invention, volume reduction can be achieved by taking into account parameters captured by the measurement. In particular, the reverberation time can represent the attack time. Thus, the acoustic properties of the room can be parameterized analogously to noise gate processing.

[0011] The invention also relates to an audio post-processing unit for post-processing an audio signal that has been captured by a microphone. Here, the reverberation properties of a room can first be determined. This is done by capturing a first audio signal as a test signal using at least one microphone located in the room. The reverberation properties of the room are determined based on the captured first audio signals from the at least one microphone. A second audio signal can be captured by the microphone in the room. The second audio signal comprises a wanted sound and a room reverberation generated in the room by the wanted sound. A prediction of a room reverberation is created based on the determined reverberation properties of the room and the captured second audio signal. The second audio signal is analyzed for locations where the room reverberation has a higher level than a wanted signal, in particular a speech signal.The volume of a second audio signal to be output is reduced at those points where the room reverberation prediction indicates that the room reverberation exceeds a threshold. In other words, the volume of the second audio signal to be output is reduced depending on the room reverberation. In particular, if the useful signal in the second audio signal has a low level while the room reverberation has a high level, the volume of the second audio signal to be output can be reduced to reduce the influence of the room reverberation on the entire second audio signal to be output.

[0012] Further embodiments of the invention are the subject of the subclaims.

[0013] Advantages and embodiments of the invention are explained in more detail below with reference to the drawing. Fig. 1 shows a schematic representation of a room with a microphone, Fig. Figure 2 shows a graph illustrating a signal and a transient level and a gain according to one aspect of the present invention, and Fig. 3 shows a block diagram of an audio post-processing system.

[0014] Fig. 1 shows a schematic representation of a room with a microphone. At least one microphone 200 is provided in a room 100, which is intended to capture a wanted signal 110. However, when capturing the wanted signal 110, the microphone 200 captures, for example, a room reverberation or reverberation 120 of the wanted sound 110 due to the acoustic properties of the room 100. The audio signal recorded by the microphone 200 thus consists of a wanted sound 110 and a room reverberation 120. The room reverberation 120 is not desired in the recorded audio signal, since only the wanted signal 110 is to be recorded.

[0015] To determine the room reverberation 120, a microphone (e.g., microphone 200) can be placed in the room 100. An audio signal can then be output as a test signal in the room 100. The microphone 200 can then record the test signal and the resulting room reverberation in order to determine the room reverberation properties of the room 100.

[0016] The captured audio signal of the microphone 200 can be post-processed in an audio processing unit 300.

[0017] As long as the level of the room reverberation 120 is lower than the level of the wanted sound 110, the influence of the reverberation sound 120 on the output signal is relatively small. Since the wanted signal 110, for example a speech signal, does not occur evenly, there may be points in the captured audio signal where the wanted sound 110 has a lower level than the room reverberation 120. In such cases, the room reverberation 120 becomes noticeable in the captured audio signal. This is undesirable. To avoid this, it is proposed to reduce the level of the output audio signal (wanted sound plus room reverberation) at those points where the room reverberation 120 is above a threshold value. This is typically the case if the level of the wanted sound 110 is lower than the level of the room reverberation 120 at that time. In these cases, the level of the audio signal (wanted signal and room reverberation) can be reduced.This reduces the influence of the room reverberation 120 on the overall audio signal.

[0018] Fig. Figure 2 shows a graph illustrating a signal and a transient level as well as a gain according to one aspect of the present invention. In Fig. 2, the signal level SL of the signal S and the transient level TL are plotted over time t. Furthermore, the gain G is plotted over time t. In particular, the gain G of the audio signal S is reduced at some points. These points correspond to the points at which the room reverberation signal would be perceptible in the overall audio signal. The detection of these points can be achieved, for example, by a prediction. This prediction of the room reverberation can be achieved by measuring the room acoustics, for example, by means of an impulse response from the microphone in the room (see Fig. 1) has been recorded.

[0019] Fig. 2 is intended to provide a better understanding of the audio post-processing used. Fig. 2 above shows the envelope of a signal S. Furthermore, the transient level TL extracted from the signal curve is shown. The transient level TL represents the current benchmark for the algorithm's prediction. If the signal S falls below a level that is determined in relation to the transient level, the existing signal S is considered reverberation (i.e., not a useful signal). Immediately afterward, a volume reduction can be applied. Fig. Figure 2 (below) shows the volume control for signal waveform S. At certain time intervals, the signal detects that only room reverberation (e.g., reverberation) is present, and the volume is reduced accordingly. If the signal level increases again, the volume is returned to a neutral level. The type and speed (i.e., "attack") of this reduction are determined by the acoustic properties of the room.

[0020] Fig.3 shows a block diagram of audio post-processing. The audio processing unit 300 has an input signal 301 and an output signal 302. The input signal 301 corresponds to the output signal 302 of the microphone 200. The audio processing unit 300 has a simulation unit 310 for simulating room reverberation based on the audio signal acquired by the microphone 200 (useful signal and room reverberation signal) and the room reverberation properties (which were acquired using a test signal and form the basis for predicting the room reverberation properties), a first volume unit 350 at the output of the simulation unit 310, and a second volume unit 360 with the audio input signal 301 as the input signal. The first and second volume units 350, 360 each serve to determine a volume. The output signal 302 of the first volume unit 350 is subtracted from the output signal 302 of the second volume unit 360 in a subtraction unit 330.In a comparison unit 340, the determined difference is compared with a threshold value.

[0021] The audio processing unit 300 further includes a gate 320 with a side input 321. The gate 320 is controlled as a function of the signal at the side input 321. Optionally, the gate 320 can decouple the output 302 from the input 301, so that no output signal 302 is output.

[0022] Optionally, the gate 320 may reduce a level of the input signal 301 if the signal at the side input 321 indicates that a level of a room reverberation is higher than a level of a useful signal. List of reference symbols 100 room 110 useful signal 120 Room Reverb 200 microphone 300 audio processing unit 301 input signal 302 Output signal 310 Simulation Unit 320 Gate 321 Side Entrance 330 Subtraction unit 340 comparison unit 350 volume unit 360 volume unit

Claims

[1] Method for audio post-processing of an audio signal captured by a microphone (200), comprising the steps: Determining reverberation properties of a room (100) by detecting a first audio signal as a test signal by at least one microphone (200) in the room (100), and Determining reverberation properties of the room (100) based on the detected first audio signals of the at least one microphone (200), detecting a second audio signal by a microphone (200) in the room (100), wherein the second audio signal comprises a useful sound and a room reverberation (120) generated by the useful sound in the room (100), Creating a prediction of a room reverberation (120) based on the determined reverberation properties of the room (100) and the detected second audio signal, analyzing the second audio signal for locations in the audio signal at which the level of the room reverberation (120) is higher than a level of a useful signal (110) in the second audio signal, Outputting the second audio signal when the level of the room reverberation (120) is lower than a level of the useful signal (110), and Outputting the second audio signal at a reduced volume when the level of the room reverberation (120) is higher than a level of a useful signal (110). [2] The method of claim 1, wherein the first audio signal represents an impulse response. [3] Audio post-processing unit for audio post-processing of an audio signal captured by a microphone (200), wherein the audio post-processing unit is designed to To determine the reverberation properties of a room (100) by a first audio signal is detected as a test signal by at least one microphone (200) in the room (100), and Reverberation properties of the room (100) are determined based on the detected first audio signals of the at least one microphone (200), and is also designed to to capture a second audio signal through a microphone (200) in the room (100), wherein the second audio signal comprises a useful sound and a room reverberation (120) generated by the useful sound in the room (100), to create a prediction of a room reverberation (120) based on the determined reverberation properties of the room (100) and the detected second audio signal, to analyze the second audio signal for locations in the audio signal at which the level of the room reverberation (120) is higher than a level of a useful signal (110) in the second audio signal, to output the second audio signal when the level of the room reverberation (120) is lower than a level of the useful signal (110), and to output the second audio signal at a reduced volume if the level of the room reverberation (120) is higher than a level of a useful signal (110).

Images