Psychoacoustic calibration of an audio playback system
Patent Information
- Application Number
- EP2024716643
- Authority / Receiving Office
- EP · EP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-03-23
- Filing Date
- 2024-03-19
- Publication Date
- 2026-01-28
Smart Images
Figure EP2024057284_26092024_PF_FP
Abstract
Description
[0001] Psychoacoustic calibration of an audio playback system
[0002] The invention relates to a method for the psychoacoustic calibration of an audio reproduction system, which comprises the steps a. Providing a masking noise with an input signal level L^, where ne N and a frequency range f M by a data processing device, generating the masking noise by a signal generator, transmitting the masking signal to an audio reproduction system and reproducing it thereon with an output sound level with unknown control deviation L e (fivi); b. Providing a test tone with an input signal level L T and a frequency f Tby the data processing device, generating the test tone by a signal generator, transmitting the test tone to the audio reproduction system and playing it back on it with an unknown output sound level; c. Indicating to a user via a suitable interface whether the test tone is perceptible or not and transmitting this information to the data processing device; d. Adjusting the input signal level of the test tone L T until the input signal level L S T of the test tone is reached, at which the test tone is just perceived by the user, whereby the input signal level L S T the masking threshold for the test tone with the frequency fr for the masking noise with the input signal level L n M and the frequency range fM; e. Storing the masking threshold L S T and calculate the relative masking threshold AL n = L n M - LS T f. Repeat steps a to e for n input signal levels L n M of the masking noise in the frequency range fM, where ne N, where the test tone has the same frequency f T g. Plot of the relative masking threshold AL n above the input signal level L n M of the masking noise; comparison of the curve with a reference curve and calculation of the control deviation L e (fM) between the input signal level and the output sound level of the masking noise from the two curves; i.e., calculating an equalization filter from the control deviations. Furthermore, a system for the psychoacoustic calibration of an audio reproduction system is disclosed, which is configured to carry out the method according to the invention.
[0003] The evaluation of audio signals is crucial in numerous applications. For example, in quality assurance in the automotive industry or in the field of household appliances; in the production of audio content such as films, music, or games; or in future big data models for noise synthesis, analogous to stable diffusion for images or ChatGPT for text.
[0004] However, audio playback systems are generally very heterogeneous in terms of the hardware used. On the other hand, audio playback system settings are constantly being changed by the user. Therefore, for many applications requiring audio playback, calibration of the playback system is necessary. Only calibration enables precise control over the sound level and frequency response of the audio playback system. Only in this way can it be ensured that the user of the respective playback system hears the sound as intended by the sound provider. This applies, for example, to the collection of auditory test subjects' judgments about sounds, for example, for quality management or as training data for training predictive models, for example, using machine learning.
[0005] Due to the enormous number of required assessments, online crowdsourcing is of particular interest. However, the heterogeneity of acoustic reproduction systems and the widespread lack of calibration hardware have so far hampered crowdsourcing. Therefore, a method is needed that universally enables the remote equalization of reproduction systems without the need for on-site access.
[0006] So far, there are essentially five approaches to remote calibration.
[0007] 1. Calibration of individual audio playback systems is carried out in laboratories using appropriate measuring technology (microphones).
[0008] However, this approach requires access to a measurement system and the technical know-how to perform the calibration. In many scenarios, such as in private households with highly heterogeneous audio playback systems, there is generally no access to measurement technology for calibration or the know-how to use it is lacking. Therefore, calibration in the laboratory, while very precise, is also associated with very high costs.
[0009] 2. The second approach attempts to identify the frequency response from a database using the type designation of the audio playback system used, e.g. the headphones, and thus compensate the frequency response.
[0010] However, identifying the entire transfer function of all conceivable audio playback systems using a single type designation is inadequate. First, the entire audio playback system usually consists of several subsystems (sound card, amplifier, loudspeakers), which would require the properties of all subsystems to be taken into account, all of which would have to be accurately recorded in a database. Second, this does not account for manufacturing-related variations in the frequency response of the playback system components. Third, it cannot be used to calibrate the playback level, since this can be freely adjusted or adjusted, for example, in the sound card or amplifier.
[0011] 3. The third approach (US 2003 / 0078515 A1) aims to allow the user to dynamically adjust the sound level of the playback system so that a defined external noise is perceived as being of the same volume, thus achieving a calibration of the overall playback level.
[0012] However, due to the lack of dedicated calibrators, no external noise sources (as in (US 2003 / 0078515 A1)) with a clearly defined sound level or spectrum are widely available. The excitation mechanisms of widely available noise sources (e.g., the rubbing of sheets of paper against each other) are highly dependent on the applied force, speed, or material type and can only produce a noise with a highly variable sound level.
[0013] 4. For the remote calibration of audio reproduction systems, psychoacoustic methods without an external reference are practical, as they require neither the use of an on-site measurement system nor the use of an external noise source with a precisely defined sound level. The fourth approach (US 8,059,833 B2) therefore attempts to allow a user to dynamically adjust the sound level of a sine tone played through the audio reproduction system so that the sound is just barely audible. The resting hearing threshold, which is known from the previously known anatomical properties of the ear regarding sensitivity, thus represents the standard against which the calibration of the audio system in use can be carried out.
[0014] The only known psychoacoustic method (US 8,059,833 B2) is based on the resting hearing threshold and thus on the anatomical properties of the ear regarding its sensitivity. This has the critical disadvantage that the resting hearing threshold must be used as an internal measurement standard. The resting hearing threshold generally increases significantly with age. However, this increase in the resting hearing threshold is also strongly dependent on the individual's previous noise exposure. Calibration based on the user's resting hearing threshold is therefore only possible with a significant error, as the measurement standard fluctuates considerably.
[0015] Except for specially soundproofed listening booths, the presence of unknown background noise must always be assumed in everyday environments where audio reproduction systems are used. This makes it impossible to distinguish whether a test sound presented represents the threshold of the quiet hearing threshold without background noise or the threshold of an unknown (side-)listening threshold in the presence of background noise. This also effectively distorts the measurement standard.
[0016] For the reasons mentioned above, this method is more suitable for compensating for individual limitations of the audio playback situation than for calibrating the audio system itself.
[0017] 5. Similar to the fourth approach, for example, in hearing aids, the playback system is assumed to be known, but the user's resting hearing threshold is assumed to be unknown. However, this can only achieve individual compensation for hearing loss. Calibration of the sound level and frequency response of an audio playback system is not intended, but rather an improvement in the audibility of sounds for the user.
[0018] Furthermore, the following methods are known from the state of the art.
[0019] US 8059833 B2 discloses a reverse psychoacoustic measurement method. This method uses the user's resting hearing threshold. However, this has the major disadvantage that background noise is usually always present, with a sound level above the resting hearing threshold, thus distorting the measurement. Furthermore, the resting hearing threshold is highly inter-individual and is subject to change, especially with increasing age.
[0020] US 2003 / 0055627 A1 deals with noise suppression in speech processing. However, the method uses masking curves only to determine the audibility of noise components, not to calibrate audio reproduction in general or its sound level specifically.
[0021] EP 0 989 776 A2 deals with the relative matching of sound levels of individual loudspeakers in a multi-channel system so that the directional information of multi-channel audio is transmitted correctly. For this purpose, the audio signals from the various loudspeakers are to be measured at the listener's position using microphones, for example. However, a microphone is not present on every audio playback system and may therefore have to be provided to implement this method, which limits the applicability of the method. Alternatively, EP 0 989 776 A2 proposes incorporating the user's perception into the loudness matching of multiple channels. However, this subjective method is not explained in detail.
[0022] Therefore, none of the approaches presented can enable remote calibration of the playback level and frequency response of an audio playback device with sufficiently high accuracy.
[0023] The present invention is therefore based on the object of providing a method that enables the precise calibration of the frequency response and playback level of any unknown audio playback system remotely using a psychoacoustic method. The method is intended to eliminate the disadvantages of the prior art and, in particular, to be insensitive to typical, low background noise levels and to individual shifts in the resting hearing threshold.
[0024] For this purpose, the invention provides a method according to claim 1 and a system according to claim 10.
[0025] Detailed description
[0026] According to the method for the psychoacoustic calibration of an audio reproduction system, a masking noise with an input signal level L^, where n e = N and a frequency range fw is first transmitted to and reproduced on an audio reproduction system by a data processing device. The masking noise is then reproduced on the audio reproduction system at an output sound level with an unknown control deviation L e (fivi). The output sound level generally does not correspond to the input signal level, since, as already described, this is influenced by the technical characteristics of the audio playback system. Calibration, in the context of the invention, means calibrating the sound pressure level and frequency response of the audio system.
[0027] A data processing device within the meaning of the invention is understood to mean, for example, a local computer or a remote server. The data processing device has suitable software to provide the masking noise and the test tone and to control the audio playback device and the signal generators accordingly.
[0028] In principle, the frequency range fw of the masking noise depends on the physical properties of the sound generation by the audio reproduction system, e.g., the driver size. In one embodiment, the frequency range fw of the masking noise is therefore limited depending on the audio reproduction system.
[0029] In one embodiment, the width of the frequency range fw of the masking noise may, for example, comprise a tone, a third, or several frequency groups.
[0030] In one embodiment, the masking noise may be, for example, a band-limited noise, a sine tone, or another noise suitable as a masking noise.
[0031] According to the invention, the frequency range fw of the masking noise lies at least partially in the audible range of the frequency spectrum. In one embodiment, the method is used for several masking noises, wherein the frequency ranges fw of the masking noises successively cover approximately the entire audible range. The frequency range of the masking noise can therefore be in the range from 20 Hz to 20 kHz. The input signal level L^ of the masking noise should be set according to the invention such that the sound level generated by the audio reproduction system lies between 20 and 100 dB, preferably between 30 and 90 dB, particularly preferably between 50 and 90 dB. These sound level ranges enable the range of the non-linear course of the curve of the relative masking threshold AL n above the input signal level L n M of the masking noise.
[0032] Furthermore, a test tone with an input signal level L T and a frequency fr are provided by the same data processing device, transmitted to the same audio reproduction system, and reproduced thereon. The test tone is also output by the audio reproduction system at an unknown output sound level, which generally does not correspond to the input signal level. The test tone can, for example, be a sine tone.
[0033] According to the invention, the masking sound and the test tone are reproduced at least partially simultaneously on the audio reproduction system. In one embodiment of the present invention, the playback of the test tone can begin before the playback of the masking sound and / or last longer than the playback of the masking sound on the audio reproduction system. The test tone can, for example, begin 40 ms, 30 ms, 20 ms, or 10 ms before the masking sound and / or be reproduced 100 ms, 200 ms, or 300 ms after the masking sound.
[0034] The audio playback system can, in principle, be any audio system that has a known relationship between input level and output level within the calibration range, preferably an approximately linear relationship. This includes, in particular, headphones and loudspeakers.
[0035] According to the invention, two signal generators are connected upstream of the audio reproduction system, which can generate the masking noise and the test tone with a variable input signal level and variable frequency or variable frequency range. In a preferred embodiment, the signal generators are integrated into the data processing device that provides the masking noise and the test tone.
[0036] The signal generators can also be integrated into a separate data processing device. For example, a data processing device located at the location of the audio playback system. In this case, a suitable connection exists to the data processing device that provides the test tone and the masking noise. A suitable connection can be a cable connection, for example, via LAN, or a wireless connection, such as a WLAN or Bluetooth connection. A wireless connection is preferred.
[0037] The audio playback system is preferably connected via a suitable connection to a data processing device having signal generators that generate the masking noise and the test tone. This is preferably the data processing device that provides the test tone and the masking noise. The masking noise and the test tone are transmitted to the audio playback device via this connection. A suitable connection can be a cable connection, for example, via LAN, or a wireless connection, such as a WLAN or Bluetooth connection. A wireless connection is preferred.
[0038] This embodiment of the method according to the invention advantageously allows the data processing device that provides the test tone and the masking noise to be located at a distance from the audio playback system. In particular, the audio playback system and the data processing device can be located anywhere and are connected to each other, for example, via the Internet.
[0039] The frequency of the test tone lies in the audible frequency range between 20 Hz and 20 kHz. According to the invention, the frequency of the test tone lies above, within, or below the frequency range fw of the masking noise, with the maximum difference between the frequency of the test tone and the frequency range of the masking noise being reached where the masking curve generated by the masking noise has dropped to the hearing threshold. In a preferred embodiment, the frequency of the test tone lies above the frequency range f^ of the masking noise. Particularly preferably, the frequency of the test tone lies in the range of the masking curve of the masking noise, in which it drops approximately linearly with the frequency.
[0040] Test tone and masking noise are generated by a user of the
[0041] The test tone is perceived by the audio playback system, with the user indicating via a suitable interface whether or not they can hear the test tone. If the sound level of the test tone generated by the audio system is too low, the test tone will be masked by the masking noise. Only above a certain sound level, which depends on the frequency of the test tone and the sound level of the masking noise generated by the audio system, can the test tone be perceived by the user despite the masking noise. This sound level of the test tone, at which the test tone is just perceptible despite the masking noise, is called the masking threshold.
[0042] A suitable interface is, for example, a computer such as a PC, tablet, or smartphone with suitable software or app. The software or app then provides an input interface through which the user can indicate whether or not the test tone is perceived.
[0043] In one embodiment of the invention, the interface is located on the data processing device used to provide the test tone and the masking noise. If the interface is not located on this data processing device, the interface is connected to the data processing device via a suitable connection. A suitable connection can be a cable connection or a wireless connection, for example via LAN, such as a WLAN or Bluetooth connection. The connection is particularly preferably a wireless connection. The information entered by the user in the interface is then transmitted to the data processing device used to provide the test tone and the masking noise.
[0044] According to the invention, the input signal level of the test tone L T adjusted until the input signal level L S Tof the test tone is reached, at which the test tone is just perceived by the user, whereby the input signal level L S T the masking threshold for the test tone with the frequency fr for the masking noise with the input signal level L n M Adjusting the input signal level of the test tone involves both increasing and decreasing the input signal level in order to adjust to the masking threshold L S T of the test tone. Adjusting the input signal level L T is carried out by the data processing device that provides the test tone and the masking noise in response to the information provided by the user as to whether or not he perceives the test tone.
[0045] Subsequently, the masking threshold L S T stored and the relative
[0046] Masking threshold AL n = L nM - L S T and then stored. The calculation of the relative masking threshold is preferably carried out on a computer, such as a PC, tablet, or smartphone using appropriate software. In a preferred embodiment, the masking threshold L S T stored on the data processing device that provides the test tone and the masking noise, and the relative masking threshold is also calculated and stored there.
[0047] The method steps already described are carried out according to the invention for n input signal levels L n M of the masking noise in the frequency range f M , where ne repeats N. The test tone with the same frequency fr is always used.
[0048] In one embodiment of the present invention, n is between 2 and 20, preferably between 3 and 10, particularly preferably between 3 and 5.
[0049] According to the invention, the course of the relative masking threshold AL n above the input signal level L n M The actual output sound level of the masking noise can be determined from the change in the resulting curve compared to a reference curve of a calibrated audio playback system. For this purpose, the resulting curve is compared with an unknown control deviation L e(fivi) is compared with a reference curve of a calibrated audio system without control deviation, and the unknown control deviation Le(fivi) between the input signal level and the output sound level of the masking noise is calculated from the two curves. The output sound level of the masking noise describes the sound level generated by the audio playback system when outputting the masking noise. According to the invention, an equalization filter is calculated from the control deviations obtained by comparing the measured curve with the reference curve. These steps preferably take place on the data processing device that provides the test tone and the masking noise.
[0050] In one embodiment, the reference curve is provided on the data processing device that generates the test tone and the masking noise. The reference curve is determined prior to application of the method according to the invention using a calibrated audio reproduction system. The calibrated audio reproduction system can be of the same design as the audio reproduction system used in the method according to the invention, but does not have to be. In one embodiment, the calibrated audio reproduction system can be calibrated using a method from the prior art, preferably using appropriate measuring technology (microphones) in a laboratory. Methods for this are known to those skilled in the art. Particularly preferably, the reference curve is determined by applying method steps a to f of the method according to the invention in the laboratory and applying the relative masking thresholds AL thus obtained. nover the input signal levels of the masking noise L^, which are identical to the sound levels of the masking noise generated by the audio reproduction system, for the calibrated audio reproduction system. According to the invention, a reference curve must be provided for each test tone with a specific frequency fr in combination with a masking noise with a frequency range fw.
[0051] In one embodiment of the invention, the calculated equalization filter is used in the reproduction of sounds by the audio reproduction system to enable calibrated audio reproduction.
[0052] In a preferred embodiment, the method according to the invention is carried out for m masking noises with m different frequency ranges fw, where me N.
[0053] If the frequency range of the masking noise is changed, the frequency of the test tone is also changed, so that the method is carried out for m test tones with m different frequencies fr, where m ≤ N. In one embodiment, m is between 1 and 50, preferably between 1 and 24, particularly preferably between 1 and 5.
[0054] If the procedure is performed for multiple masking noises with m different frequency ranges fw, an equalization filter can be calculated for the frequency range covered by the frequencies of the m masking noises. The resulting equalization filter enables calibrated audio reproduction of the audio playback system across a corresponding frequency range. The frequency range is determined by the frequency ranges of the m masking noises. The frequency response of an audio playback device can thus be determined.
[0055] The invention further provides a system which is configured to run the method according to the invention. The system comprises an audio reproduction system, at least one computing unit, an interface for a user, and two signal generators. In particular, the system according to the invention is configured to run a method for the psychoacoustic calibration of an audio reproduction system, comprising the steps of: a. Providing a masking noise with an input signal level L 1 ^, where ne N and a frequency range f M by a data processing device, generating the masking noise by a signal generator, transmitting the masking noise to an audio reproduction system and reproducing it thereon with an output sound level with unknown control deviation L e (fivi); b. Providing a test tone with an input signal level L T and a frequency fT by the data processing device, generating the test tone by a signal generator, transmitting the test tone to the audio reproduction system and playing it back on it with an unknown output sound level; c. Indicating to a user via a suitable interface whether the test tone is perceptible or not and transmitting this information to the data processing device; d. Adjusting the input signal level of the test tone L T until the input signal level L S T of the test tone is reached, at which the test tone is just perceived by the user, whereby the input signal level L S T the masking threshold for the test tone with the frequency fr for the masking noise with the input signal level L n M and the frequency range fM; e. Storing the masking threshold L S T and calculate the relative
[0056] Masking threshold ALn = L n M - L S T f. Repeat steps a to e for n input signal levels L n M of the masking noise in the frequency range fM, where ne N, where the test tone has the same frequency f T g. Plot of the relative masking threshold AL n above the input signal level L n M of the masking noise; comparison of the curve with a reference curve and calculation of the control deviation L e (fM) between the input signal level and the output sound level of the masking noise from the two curves; i.e., calculation of an equalization filter from the control deviations; can be performed thereon. All features described for the method according to the invention also apply to the system according to the invention, and vice versa.
[0057] For the remote calibration of audio reproduction systems, psychoacoustic methods without an external reference are advantageous because they require neither the use of an on-site measurement system nor the use of an external noise source with a precisely defined level. When using psychoacoustic methods in the study of hearing properties, the audio reproduction system is always assumed to be calibrated, i.e., known, and the properties of the ear are unknown. However, for the calibration of audio reproduction systems using psychoacoustic methods, the properties of the audio reproduction system are assumed to be unknown and the properties of the ear are known. The only known psychoacoustic calibration method is based on the resting hearing threshold and is highly error-prone due to the fluctuation of the measurement standard resulting from the individual variance in the resting hearing threshold and the unknown background noise.This makes the process more suitable for compensating for individual limitations of the playback situation than for equalizing the audio system for controlled playback.
[0058] Audio reproduction systems are based almost exclusively on electrodynamic loudspeakers. A key property is their linear behavior below very high playback levels, which can be assumed to be known. The property of frequency selectivity (i.e., masking) is used, for example, in MP3 coding. This exploits the property that a high-level masking noise masks a test tone of higher frequency but lower level. The level difference between the masking noise and the test tone, i.e., the relative masking threshold at which the test tone is still detectable, is assumed to be linear.However, this is not true, since the relative masking threshold decreases with increasing sound level of the masking noise [1]. Advantageously, the present invention can make the non-linear course of the relative masking threshold technically usable as a measurement standard in a psychoacoustic calibration method for identifying the system properties (output sound level, frequency response) of an audio reproduction system.
[0059] Using two signal generators connected upstream of the unknown audio system, a masking noise and a test tone with variable sound levels can be generated, thus stimulating the audio playback system. The user's ability to detect the test tone represents the measurement standard in the form of the relative masking threshold. Since the audio playback system can be assumed to be linear, both the playback levels of the masking noise and the test tone are unknown, but the level difference between the masking noise and the test tone remains constant, despite the change in the overall playback level caused by the unknown audio playback system. Thus, the relative masking threshold depends exclusively on the user's ability to detect the test tone.
[0060] The mean-free curve of the relative masking threshold over the sound level of the masking noise is subject to only slight inter-individual fluctuations. Therefore, the actual reproduced output sound level of the masking noise can be determined from the deviation of the relative masking threshold curve of a specific user from a reference measurement determined in advance on a calibrated and thus equalized audio playback system. The relationship between the input signal level and the reproduced output sound level can ultimately be used to create a digital filter, which is then used to equalize the audio playback system.
[0061] Thus, the method according to the invention can advantageously utilize ear characteristics that have not previously been used as measurement standards for calibration procedures. Furthermore, the method according to the invention does not require any external reference noise sources or a measurement system, in particular a microphone, for calibration. In contrast to hearing threshold-based psychoacoustic calibration methods, the method is insensitive to individual fluctuations in the hearing threshold. Since playback levels well above the hearing threshold can be used, the method is insensitive to background noise close to the hearing threshold, which is always encountered in everyday playback situations.
[0062] The invention is explained in more detail below with reference to 4 figures and 1 embodiment.
[0063] Figure 1 (A) shows an example of the input signal level of Lydes test tone over the
[0064] Frequency fr of the test tone; (B) shows an example of the course of the relative masking threshold AL n above the input signal level L n M of the masking noise;
[0065] Figure 2 shows a flow diagram of the method according to the invention; Figure 3 (A) shows by way of example the input signal level Ly of the test tone over the
[0066] Frequency fr of the test tone; (B) shows an example of the course of the relative masking threshold AL n above the input signal level of the masking noise and the course of the masking threshold L Re f from the reference measurement (reference curve);
[0067] Figure 4 (A) to (C) shows an example of a measurement procedure.
[0068] Figure 1 (A) shows, as an example, the input signal level LT of the test tone versus the frequency fr of the test tone. The figure also shows the hearing threshold and the masking curves generated by the masking noises with different input signal levels LM. The drop in the masking curve becomes smaller with increasing masking level. In this example, the masking noise was a band-limited noise with a constant frequency range f. M of a third octave and a center frequency of 1 kHz. Bars 10 and 20 represent two test tones with the same frequency fr, but different input signal levels LT. To better distinguish test tones 10 and 20, they are shown side by side.
[0069] According to the invention, for example, the test tone 10 and a masking noise with an input signal level = 80 dB is output via the audio playback system. The input signal level LT of the test tone is now adjusted until the test tone is just perceptible to a user. This corresponds to the input signal level of the test tone at which bar 10 intersects the masking noise curve with an input signal level L^ = 80 dB. This input signal level of the test tone corresponds to the masking threshold L S T . From the masking threshold L S T and the input signal level L^ = 80 dB of the masking noise is the relative masking threshold AL n = L^ - L S T calculated.
[0070] Subsequently, a masking noise with a different input signal level, for example L^ = 90 dB, is played back on the audio playback device together with the test tone 20. The frequency range of the masking noise fM and the frequency of the test tone remains unchanged. Again, the input signal level LT of the test tone is adjusted until the test tone is just perceptible to a user. From the masking threshold L thus determined, S T and the sound level of the masking noise, the relative masking threshold can be calculated. This can be done for n different input signal levels of the masking noise. Figure 1 (B) shows an example of the curve of the relative masking threshold AL. n , ie the difference between the input signal level of the masking noise and the masking threshold of the test tone above the input signal level L n M The masking noise with a constant frequency range fw is plotted for a test tone of a constant frequency fr. The nonlinear curve is clearly visible.
[0071] Figure 2 shows a flow chart of the method according to the invention. The diagram represents both the calibration of an audio reproduction system by the method according to the invention and the reproduction of audio signals by the then calibrated audio reproduction system.
[0072] First, a masking noise 204 and a test tone 101 are provided for calibration by a data processing device 400. These are generated by signal generators and reproduced by an audio playback system 40 with unknown properties. A user of the audio playback system 40 is thus presented with the masking noise and the test tone 60. The sound level of ambient noise 50 can be above the hearing threshold, but advantageously below the output sound level of the test tone and the masking noise generated by the audio system, so that ambient noise has no influence on the inventive method. A user with an unknown hearing threshold then indicates via a suitable interface 80 whether the test tone is audible or not 70. If the test tone is not audible, the input signal level of the test tone is changed 100.This loop is repeated until the masking threshold for the test tone is reached. In this case, the masking threshold is stored (200) and the relative masking threshold is calculated (201). These steps are repeated for different input signal levels of the masking noise (302, 203), resulting in the curve of the relative masking threshold AL. n above the input signal level L n M of the masking noise can be determined. This curve is compared with a reference curve 300 and the deviations are used to calculate a level correction for the current frequency range f Mof the masking noise 310 is calculated. The method is then advantageously run through 311, 312 for m masking noises with m different frequency ranges and m test tones with different frequencies. From the level corrections of the m frequency ranges of the masking noises, a transfer function of the audio system can be formed, and thus an inverse filter for equalization 320 can be calculated. When an audio signal 30 is reproduced, the inverse filter for equalization 320 cancels out the frequency response of the audio reproduction system, and the audio reproduction from the audio reproduction system 40 is calibrated.
[0073] Box 500 identifies all process steps that take place on-site at the audio playback system. Box 400 identifies the process steps that are location-independent, meaning they can take place both at the audio playback system's location and at a remote location.
[0074] Figure 3 (A) illustrates, by way of example, the input signal level L of the test tone over the frequency fr of the test tone. The figure shows three masking noises 1a, 2a, and 3a with different sound levels L^ (n = 1, 2, 3) and their generated masking curves 1c, 2c, 3c. For a test tone of a frequency fr, the masking threshold L is determined for all three masking noises one after the other using the method according to the invention. S T (indicated in the figure with the reference symbols 1b, 2b, 3b) are determined. From the masking thresholds L S T and the input signal levels L n M of the masking noises, the relative masking threshold AL n = L n M - L S T calculated.
[0075] Figure 3 (B) shows an example of the course of the relative masking threshold AL n above the input signal level L n Mof the masking noise and a reference curve AL fie ^(1d, 2d, 3d). The reference curve was obtained by measuring with a calibrated audio playback system. From the deviation of the curve for AL n and the curve for AL fie ^can be a level correction for the frequency range f M of the masking noise. An equalization filter can be calculated from the level corrections for several frequency ranges fM of the masking noise.
[0076] Figure 4 (A) to (C) shows an example of a measurement sequence and is explained in more detail in Example 1.
[0077] Example 1
[0078] Figure 4 (A) shows the input signal level LT for a test tone over the frequency fr of the test tone. The method according to the invention was carried out for m=2 frequency ranges of the masking noise as well as test tones with m=2 frequencies. A band-limited noise was used as the masking noise and a sine tone as the test tone. Figure 4 shows the measurement results for m=2 frequency ranges of the masking noise and m=2 frequencies of the test tones as examples. The method was first carried out for the masking noise m=1 with a frequency range f M of a third octave and a center frequency of 1 kHz and a test tone with the frequency f T = 1 .7 kHz. The input signal of the masking noise was reproduced by the audio system by an unknown control deviation L e(fivi) and output with an output sound level that did not correspond to the input signal level. The input signal level of the test tone LT was varied until the masking threshold at L S T was found. The course of the to determine the unknown control deviation L e (fivi) at fM shifted input signal level of the masking noise, relative masking threshold compared to a reference curve is shown in Figure 4 (B). For the center frequency of 1 kHz, a control deviation L e (fM) of +10 dB of the actual sound level generated by the masking noise from the input signal level of the masking noise L n M Subsequently, the procedure steps were carried out for the masking noise m=2 with a frequency range fM of a third octave and a center frequency of 2 kHz as well as a test tone m=2 with a frequency f T= 3.4 kHz. The course of the to determine the unknown control deviation L e (fM) at fM shifted input signal level of the masking noise, relative masking threshold compared to a reference curve is shown in Figure 4 (C). For the center frequency of 2 kHz, a control deviation L e (fM) of -10 dB of the actual sound level generated by the masking noise from the input signal level of the masking noise L n M For each frequency range fM of the masking noise, the procedure was applied for at least n = 3 sound levels L n M of the masking noise. From the determined control deviations L e(fM) of all frequency ranges of the masking noise, each calculated between the curves of the relative masking thresholds and the reference curves, an equalization filter was finally calculated over the used frequency ranges of the masking noise.
[0079] When audio signals were later played back on the audio playback device, the equalization filter was applied, and a calibrated audio signal could be output. References
[0080] [1] R. J. Bakerand S. Rosen, "Auditory filter nonlinearity across frequencies using simultaneous notched-noise masking," The Journal of the Acoustical Society of America, vol. 119, no. 1, pp. 454-462, 2006.
[0081] List of reference symbols la, 2a, 3a masking noise lb, 2b, 3b relative masking threshold lc, 2c, 3c masking curve
[0082] 1 d, 2d, 3d reference curve
[0083] 10, 20 test tone
[0084] 30 audio signal
[0085] 40 Audio playback system
[0086] 50 ambient sounds
[0087] 70 test tone audible
[0088] 80 interface
[0089] 100 Change in the input signal level of the test tone
[0090] 101 Test tone
[0091] 200 Saving the masking threshold
[0092] 201 Calculating the relative masking threshold
[0093] 204 Masking noise
[0094] 300 Reference curve
[0095] 310 Level correction
[0096] 320 equalization filters
[0097] 400 location-independent process steps
[0098] 500 procedural steps on site at the audio system
Claims
Claims 1 . A method for the psychoacoustic calibration of an audio reproduction system (40), comprising the steps of a. providing a masking noise (204) with an input signal level L^, where n e N and a frequency range f M by a data processing device, generating the masking noise (204) by a signal generator, transmitting the masking noise (204) to an audio reproduction system (40) and reproducing it thereon with an output sound level with unknown control deviation L e (fivi); b. Providing a test tone (101) with an input signal level L Tand a frequency by the data processing device, generating the test tone (101) by a signal generator, transmitting the test tone (101) to the audio reproduction system (40) and reproducing it thereon with an unknown output sound level; c. Indicating to a user via a suitable interface (80) whether the test tone (101) is perceptible or not and transmitting this information to the data processing device; d. Adjusting the input signal level of the test tone L T until the input signal level L S T of the test tone (101) is reached, at which the test tone (101) is just perceived by the user, whereby the input signal level L S T the masking threshold for the test tone (101) with the frequency fr for the masking noise with the input signal level L n M and the frequency range fM; e. Storing the masking threshold L S Tand calculate the relative masking threshold AL n = L n M - L s r f. Repeat steps a. to e. for n input signal levels L n M of the masking noise (204) in the frequency range fM, where ne N, where the test tone (101) has the same frequency fr in each case; g. Plotting the course of the relative masking threshold AL n above the input signal level L n M of the masking noise (204); comparing the curve with a reference curve (300) and calculating the control deviation Le(fM) between the input signal level and the output sound level of the masking noise (204) from the two curves; h. calculating an equalization filter (320) from the control deviations.
2. Method according to claim 1, characterized in that the method is carried out for m masking noises (204) with m different frequency ranges fw, where me N.
3. Process according to one of the preceding claims, characterized in that n is between 2 and 20, preferably between 3 and 10, particularly preferably between 3 and 5.
4. Method according to one of the preceding claims, characterized in that the masking noise (204) is a band-limited noise, a sine tone or another noise.
5. Method according to one of the preceding claims, characterized in that the frequency range f^of the masking noise (204) at least partially covers the audible range of the frequency spectrum.
6. Method according to one of the preceding claims, characterized in that the input signal level L n M is set so that the sound level generated by the audio reproduction system (40) is between 20 and 100 dB, preferably between 30 and 90 dB, particularly preferably between 50 and 90 dB.
7. Method according to one of the preceding claims, characterized in that the reference curve (300) was determined by a calibrated audio reproduction system (40).
8. Method according to one of the preceding claims, characterized in that the audio playback device (40) is a loudspeaker or headphones.
9. Method according to one of the preceding claims, characterized in that the calculated equalization filter (320) is used in the reproduction of sounds by the audio reproduction system (40) in order to enable a calibrated audio reproduction.
10. A system for the psychoacoustic calibration of an audio reproduction system (40), configured to provide a method for the psychoacoustic calibration of an audio reproduction system (40), comprising the steps of a. Providing a masking noise (204) with an input signal level L^, where n e = N and a frequency range f Mby a data processing device, generating the masking noise (204) by a signal generator, transmitting the masking noise (204) to an audio reproduction system (40) and reproducing it thereon with an output sound level with unknown control deviation L e (fivi); b. Providing a test tone (101) with an input signal level L T and a frequency by the data processing device, generating the test tone (101) by a signal generator, transmitting the test tone (101) to the audio reproduction system (40) and reproducing it thereon with an unknown output sound level; c. Indicating to a user via a suitable interface (80) whether the test tone (101) is perceptible or not and transmitting this information to the data processing device; d. Adjusting the input signal level of the test tone L T until the input signal level L S Tof the test tone (101) is reached, at which the test tone (101) is just perceived by the user, whereby the input signal level L S T the masking threshold for the test tone (101) with the frequency fr for the masking noise with the input signal level L n M and the frequency range f M is; e. Saving the masking threshold L S T and calculate the relative masking threshold AL n = L n M - L s r f. Repeat steps a. to e. for n input signal levels L n M of the masking noise (204) in the frequency range fM, where ne N, where the test tone (101) has the same frequency fr in each case; g. Plotting the course of the relative masking threshold AL n above the input signal level L n Mof the masking noise (204); comparing the curve with a reference curve (300) and calculating the control deviation Le(fM) between the input signal level and the output sound level of the masking noise (204) from the two curves; h. calculating an equalization filter (320) from the control deviations; can run on it.