GB2583438A - Signal processing device for headphones - Google Patents

Abstract

A method of processing stereo sound comprising receiving a left input signal L and a right input signal R. A first transfer function H1 is applied to the left and right input signals to generate left L’ and right R’ intermediate signals. A second transfer function H2 is applied to the left and right intermediate signals to generate left L’’ and right R’’ output signals, which are output to the corresponding transducers in headphones. Each of the left and right output signals is generated as a weighted combination of at least one of the right and left input signals, at least one of the right and left input signals modified by the first transfer function, at least one of the right and left input signals modified by the second transfer function, and at least one of the right and left input signals modified by both the first and second transfer functions. Preferably the first transfer function is improves depth perception, and the second transfer function provides lateral localisation of individual voices.

Description

Signal processing device for headphones

Field

[0001] We describe a signal processing device and associated method for processing audio sound for binaural devices such as headphones or earphones.

Background

[0002] Commercial audio content is typically designed to be played on loudspeakers and the content is often created to provide a flat response for conventional loudspeaker set-ups such as that shown in Figure la. A flat response is one in which the input sound is reproduced to the user as an output sound without enhancements or changes. The aim is to create the impression of sound heard from various directions, as in natural hearing. As set out in a paper entitled "Optimum Reproduction Matrices for Multispeaker Stereo" by Gerzon published in 1992, there is no consensus as to the ideal type of stereo recording technique. In general, Stereophonic recordings are produced with a variety of microphone techniques, for example using just two microphones or an array of microphones. The finished recording is delivered commercially as two channels for loudspeakers.

[0003] Figure la shows a typical listening arrangement in which a pair of loudspeakers 1, 2 are arranged in a room having room boundaries 7. As shown, the sound in each audio channel travels from the respective loudspeaker 1, 2 to the listener's head 3. There are several paths for the sound to travel from the loudspeaker to a user. There are ipsilateral paths 4 which are direct paths from a loudspeaker to the user's ear which is closest (i.e. on the same side) to the loudspeaker. There are also contralateral paths 5 which are direct paths from the loudspeaker to the user's ear which is further (i.e. on the opposite side) from the loudspeaker. There are also indirect paths 6 which are paths from each loudspeaker which reach a user's ear via reflection from the room boundaries. It will be appreciated that for simplicity the illustrated paths represent just a small number of the available paths. For example, sound will produce reverberation as it reflects from more than one boundary and travels around the room.

[0004] When a user listens to a music recording via headphones (which may be termed a binaural arrangement), the sound travels direct from one headphone to the adjacent ear. Sound does not reach a user's ear via reflections from the room nor from the headphone adjacent the other ear. Moreover, the ipsilateral paths from a headphone to a user's ear are considerably shorter than the ipsilateral paths from a loudspeaker to a user's ear on the same side. Accordingly, stereophonic content which has been optimised for the conventional loudspeaker set-up produces a very different sound experience for the headphone user to that for the loudspeaker user who also receives sound along contralateral and indirect paths.

[0005] The ipsilateral paths 4 and the contralateral paths 5 have associated transfer functions. The term transfer function is generally preferred to frequency response because it captures more than just the amplitude response. The term transfer function may be considered equivalent to the impulse response, the combined phase and amplitude response, or the complex frequency response. These transfer functions are known as Head Related Transfer Function (HRTF).

[0006] Various studies comparing the experience for headphone and loudspeaker users have been undertaken. For example, "Listener Preferences for In-Room Loudspeaker and Headphone Target Responses" by Olive et al published at AES 135th Convention, 2013 October 17-20 describes how listeners adjusted the relative bass and treble levels of three music programs reproduced through loudspeakers and headphones respectively. Figure 1 b illustrates a target frequency response which is disclosed in this paper and which illustrates that a flat response may not be the most appropriate frequency response for a headphone. [0007] The problem of reproducing program material that has been recorded stereophonically in a binaural system is also recognised in US3088997. This explains that feeding left and right stereophonic channel signals separately to each of left and right earphones leads to extreme left and right signals which appear to originate directly outside the listener's ears, resulting in a gross distortion in space perspective. This is addressed in US3088997 by modifying the signal in each channel and cross-feeding energy between the channels so as to simulate the sound pressure conditions prevailing at the ears of a listener located in front of a pair of spaced loudspeakers in a stereophonic array.

[0008] Other variations on earphone or headphone sound production are also described in the prior art. For example, US3962543 and US6021206 describe methods for controlling acoustical output for earphones or headphones in response to rotation or other movement of a listener's head. US6259795 describes how sound can be distributed to multiple headphone users. US6741706 describes efficient convolution of input audio signals with impulse response functions to create "colour" in the audio signal for a headphone user. US9933989 describes the use of metadata to improve binaural rendering in headphones. US6178245 describes how to generate a multi-channel output audio signal from a single-channel input audio signal and the particular orientation of a user's head. US7536021 describes the use of filtering effects in stereo headphone devices to enhance spatialization of source around a listener. US2018/003522331 describes how to generate reverberation for headphone virtualization.

[0009] The present applicant has recognised the need for an improved method of delivering content which has been optimised for delivery through loudspeakers to headphones.

Summary

[0010] According to the present invention there is provided an apparatus and method as set forth in the appended claims. Other features of the invention will be apparent from the dependent claims, and the description which follows.

[0011] We describe a method of processing stereo sound for headphones, the method comprising receiving a left input signal and a right input signal, applying a first transfer function to the left and right input signals to generate left and right intermediate signals, applying a second transfer function to the left and right intermediate signals to generate left and right output signals; and outputting the left output signal to a left transducer in the headphones and the right output signal to a right transducer in the headphones; wherein each of the left and right output signals is generated as a weighted combination of at least one of the right input signal and the left input signal, at least one of the right input signal and the left input signal modified by the first transfer function, at least one of the right input signal and the left input signal modified by the second transfer function and at least one of the right input signal and the left input signal modified by both the first transfer function and the second transfer function. [0012] The left and right transducers may also be termed left and right loudspeaker drive units (the terms may be used interchangeably). Each transducer may be a component within a loudspeaker, e.g. a complete device such as a cabinet comprising one or more transducers.

The left and right input signals may be derived from multi-channel content, for example by using a down mix method (e.g. a decoder setting) to obtain two channels for two loudspeakers from the multi-channel content.

[0013] The method thus comprises a first stage in which a first transfer function is used and a second stage in which a second transfer function is used. A transfer function may be defined as a function which relates the output of the function (i.e. the intermediate or output signal) to the input signal. The transfer function may be implemented by a filter circuit and/or a delay line. The first and second transfer functions may thus be selected to apply different and complementary effects to the input signals; which may have been received from any suitable sound source. For example, the first transfer function may be selected to improve a user's perception of depth of sounds within the stereo sound, e.g. to provide a better impression of front-back soundstage depth to a user. The second transfer function may be selected to provide better lateral localization of individual voices, e.g. of instruments and singers, within the stereo sound when listening through the headphones. Such transfer functions may thus improve the user experience when listening to input signals which have been optimised for loudspeaker playback rather than headphone use. For example, the second transfer function may be equivalent to a first order low-pass filter at 700Hz with pass-band gain which is less than 1.

[0014] Applying the first transfer function to generate the left intermediate signal may comprise subtracting the right input signal modified by the first transfer function from the combination of the left input signal and the left input signal modified by the first transfer function. Similarly, applying the first transfer function to generate the right intermediate output may comprise subtracting the left input signal modified by the first transfer function from the combination of the right input signal and the right input signal modified by the first transfer function. The first stage may thus be considered to be a difference cross-feed function. In other words, each of the right and left intermediate signals is a weighted combination of the corresponding (i.e. same side) input signal and the opposite side input signal wherein the weights are 1 for each of the corresponding input signal and the corresponding input signal modified by the transfer function and -1 for the opposite side signal modified by the transfer function. It will be appreciated that a negative weighting means that the signal is effectively subtracted.

[0015] The weighted combination for the left output signal may comprises a weight of 1 for each of the left input signal, the left input signal modified by the first transfer function and the right input signal modified by the second transfer function and a weight of -1 for the right input signal modified by the first transfer function. Similarly, the weighted combination for the right output signal may comprise a weight of 1 for each of the right input signal, the right input signal modified by the first transfer function and the left input signal modified by the second transfer function and a weight of -1 for the left input signal modified by the first transfer function.

[0016] In one arrangement, the second stage may also be a difference cross-feed function.

Applying the second transfer function to generate the left output signal may comprise subtracting the left intermediate signal modified by the second transfer function from the combination of the left intermediate signal and the right intermediate signal modified by the second transfer function. Similarly, applying the second transfer function to generate the right output signal may comprise subtracting the right intermediate signal modified by the second transfer function from the combination of the right intermediate signal and the left intermediate signal modified by the second transfer function.

[0017] The first and second stages may thus be considered to be difference cross-feed functions with the differences reversed relative to each other. Applying the first and second transfer functions in this way means that the first and second transfer functions do not cancel each other out. However, if we consider the central sounds in both channels represented by L+R, the transfer function for this signal condition is 1, i.e. adding the two weighted sums means that all the terms cancel out except for the original L and R signals. Accordingly, the central sounds are unaltered by the two stage signal processing.

[0018] In such a double difference cross-feed arrangement, the weighted combination for the left output signal may comprise a weight of -1 for the left input signal modified by the second transfer function, a weight of 2 for the right input signal modified by both the first and second transfer functions and a weight of -2 for the left input signal modified by both the first and second transfer functions. Similarly, the weighted combination for the right output signal may comprise a weight of -1 for the right input signal modified by the second transfer function, a weight of 2 for the left input signal modified by both the first and second transfer functions and a weight of -2 for the right input signal modified by both the first and second transfer functions. [0019] In other words, the left output signal may be generated from: L"= L+ LH1+ RH2+ 2RH, H2 -RHO -LH2-2LH,H2 where L" is the left output signal, L is the left input signal, R is the right input signal, H, is the first transfer function and H2 is the second transfer function. Similarly, the right output signal may be generated from: R"= R + RH, + LH2+ 2LH, H2-LH, -RH2 -2RH,H2 where R" is the right output signal, L is the left input signal, R is the right input signal, H1 is the first transfer function and H2 is the second transfer function.

[0020] The method may further comprise applying a gain effect when applying the second transfer function. It will also be appreciated that the gain effect could be applied when applying the first transfer function depending on the nature of the first and second transfer functions which have been selected. The gain effect may be a fixed constant having a value between 0 and 1, more preferably in the range 0.4 to 0.8 (i.e. for 70 to 80% correlation). The fixed constant value may be selected to achieve a desired effect, e.g. based on correlation data for the left and right input signals. For example, the gain effect G may be calculated from: 2c -100 G= where c is the percentage correlation that is expected in the music recording.

[0021] Where a gain effect is applied, the weighted combination for the left output signal may comprises a weight of -G for the left input signal modified by the second transfer function, a weight of 1+G for the right input signal modified by both the first and second transfer functions and a weight of -1-G for the left input signal modified by both the first and second transfer functions. Similarly, the weighted combination for the right output signal may comprise a weight of -G for the left input signal modified by the second transfer function, a weight of 1+G for the left input signal modified by both the first and second transfer functions and a weight of -1-G for the right input signal modified by both the first and second transfer functions.

[0022] Applying the second transfer function and the gain effect may comprise calculating the left output signal by subtracting the left intermediate signal modified by both the gain setting and the second transfer function from the total of the left intermediate signal summed with the right intermediate signal modified by the second transfer function and calculating the right output signal by subtracting the right intermediate signal modified by both the gain setting and the second transfer function from the total of the right intermediate signal summed with the left intermediate signal modified by the second transfer function. The gain setting may be applied before applying the second transfer function.

[0023] In other words, the left output signal may be generated from: L"= L+ Lfrli+ RH2+ (1+G)RHI H2 RHI -GLH2-(1+G)LH1H2 where L" is the left output signal, L is the left input signal, R is the right input signal, H, is the first transfer function and H2 is the second transfer function. Similarly, the right output signal may be generated from: R"= R + RH1 + LH2+ (1+G)L1-11 H2 -GRH2 -(1+G)RH1H2 where R" is the right output signal, L is the left input signal, R is the right input signal, H1 is the first transfer function and H2 is the second transfer function.

[0024] We also describe a non-transitory computer readable medium comprising program instructions which when carried out on a device cause the device to carry out the method described above.

[0025] We also describe a signal processing device comprising: a processor which is configured to carry out the method described above. For example, the processor is configured to receive a left input signal and a right input signal, apply a first transfer function to the left and right input signals to generate left and right intermediate signals, apply a second transfer function to the left and right intermediate signals to generate left and right output signals; and output the left output signal to a left transducer in the headphones and the right output signal to a right transducer in the headphones. Each of the left and right output signals may be generated as a weighted combination of at least one of the right input signal and the left input signal, at least one of the right input signal and the left input signal modified by the first transfer function, at least one of the right input signal and the left input signal modified by the second transfer function and at least one of the right input signal and the left input signal modified by both the first transfer function and the second transfer function.

[0026] Although a few preferred embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes and modifications might be made without departing from the scope of the invention, as defined in the appended claims.

Brief description of drawings

[0027] For a better understanding of the invention, and to show how embodiments of the same may be carried into effect, reference will now be made, by way of example only, to the accompanying diagrammatic drawings in which: [0028] Figure la is a schematic illustration of a user listening to loudspeakers; [0029] Figure 1 b is a graph of a target frequency response for a pair of headphones showing the variation in sound pressure level (dB) against frequency (Hz); [0030] Figure 2a is a schematic illustration of a difference cross-feed method; [0031] Figure 2b is a schematic illustration of a direct cross-feed method; [0032] Figure 3a is a schematic illustration of a method of processing two channel stereophonic sound for binaural use; [0033] Figure 3b is an alternative schematic illustration of a method of processing two channel stereophonic sound for binaural use; [0034] Figure 4 is a schematic illustration of an alternative second stage which can be used in the method of Figure 3a; [0035] Figure 5 is a flowchart illustrating the steps of Figures 3a to 4; [0036] Figure 6 is a schematic block diagram illustrating the components of a system for implementing the method of any one of Figures 3a to 5; [0037] Figure 7a is a graph plotting the magnitude (dB) against frequency (Hz) for a particular transfer function; [0038] Figure 7b is a graph plotting the magnitude (dB) against frequency (Hz) for the outputs after the transfer function of Figure 7a has been applied to the left and right intermediate signals at a second stage as shown in Figure 3a; [0039] Figure 7c is a graph plotting the magnitude (dB) against frequency (Hz) for the outputs after the transfer function of Figure 7a has been applied to the left and right channels at a first stage as shown in Figure 3a; [0040] Figures 8a and 8b are graphs of magnitude (dB) against frequency (Hz) showing the difference to the output for different values of G as in Figure 4.

Detailed description of drawings

[0041] As set out above, the difficulties of reproducing stereophonic material in a binaural environment are recognised in the literature. A schematic representation of a difference cross-feed method (e.g. one similar to that applied in US3088997) is shown in Figure 2a. The input left and right stereophonic signals L, R and the output left and right binaural signals L', R' are connected via a difference cross-feed transfer function labelled Hx.

[0042] The table below summarises the effect of the processing on the left output for the following notional signals when the sound is recorded stereophonically: * A signal L (where L#0 & R=0) which represents a sound placed to the left (same side) of the sound stage * A signal R (where R#0 & L=0) which represents a sound placed to the right (opposite side) of the sound stage * A signal L + R (where L=R#0) which represents a centrally placed sound; and * A signal L -R (where L=-R#0) which may represent reverberation (e.g. sound along the indirect paths similar to those in Figure la) which occurred between the microphone and the sound source Signal Condition Transfer function L->L' 1-Hx R->L' Hx (L+R)->L' 1 (L-R)->12 1-2Hx [0043] Analysis of the right output follows mutatis mutandis. A real-world recording may be considered to be a superposition of such signal types that the listener subconsciously decomposes and interprets. The table shows that with such an arrangement as suggested by US3088997, the timbre of a central sound is unaffected. Furthermore, the table shows that cross-feed transfer function may be freely chosen to achieve the desired effect on the R->L' signal condition. However, the cross-feed transfer function may then adversely affect both the L->L' and (L-R)->L' signal conditions. The reduction in the L-R signal leads to a loss of reverberation information which in turn leads to a loss of image depth (i.e. loss of front-back soundstage depth).

[0044] A simple alternative to the difference cross-feed method described in relation to Figure 2a is the direct cross-feed method which is schematically illustrated in Figure 2b. As in Figure 2a, there are input left and right stereophonic signals L, R and output left and right binaural signals L', R'. The transfer functions for the ipsilateral paths (i.e. paths 4 in Figure la) are simulated by the direct transfer function Hd and the transfer functions for the contralateral paths (i.e. paths 5 in Figure 1 a) are simulated by the direct transfer function Hd. Paths 5 are longer than paths 4 so Hx typically includes delay with respect to Hd.

[0045] The table below summarises the effect of the processing on the left output for the following notional signals when the sound is recorded stereophonically. As before the analysis of the right output follow mutatis mutandis: Signal Condition Transfer function L-*L' Hd R->L' Hx (L+R)->L' Hx + hid (L-R)->L' Hx -Hd As explained above, the most important sounds are usually centrally placed and thus ideally Hx + Hd approximates to 1 in order that the timbre of centrally placed sounds is unaffected.

This is difficult to achieve especially when Hx has greater delay than Hd because Hx represents the longer path to the opposite ear.

[0046] Figure 3a illustrates a double difference cross-feed method which can be implemented in an aspect of the present techniques. As noted above, a single difference cross-feed method does not affect a central sound because the effect of the transfer function is effectively cancelled out. Thus, the cross-feed transfer function may be freely chosen to achieve the desired effect on the signal condition. Combining two such difference cross-feed methods in sequence, i.e. in two stages, will also have no impact on the central sound. Accordingly, a first transfer function H1 may be applied in the first stage to achieve a first desired effect on the (L-R)->L' signal condition, e.g. to improve image depth -to improve front-back soundstage depth. Thereafter a second transfer function H2 may be applied in the second stage to achieve a second desired effect on the R'->L" signal condition, e.g. to improve lateral localization, e.g. of instruments and voices within the recording. The first transfer function typically involves no intentional signal delay whereas the second transfer function typically simulates the path from a loudspeaker to the opposite side ear (i.e. the contralateral path) and does typically have intentional delay. The transfer functions are explained in more detail below. However, it is noted that H1 may have a positive constant value, i.e. H1 >0. Furthermore, in some examples, H, may be similar to H2 but with an adjusted overall gain. By contrast, H2 must be designed well to result in the desired improvements.

[0047] As shown in Figure 3a, there are input left and right stereophonic signals L, R and output left and right binaural signals L", R". The first stage of processing results in intermediate output left and right signals L', R' which are the input signals in the second stage of processing. Considering the first stage separately, the intermediate output left and right signals L', R' are determined by: L'=L+(L-R)11,=L+LI11-RH, R'=R-(L-R)1-11=R+RH,-LH, In other words, the left intermediate output is the sum of the left input signal and the difference between the left and right input signals modified by the first transfer function, i.e. the left intermediate output is calculated by subtracting the right input signal modified by the first transfer function from the total of the left input signal summed with the left input signal modified by the first transfer function. Similarly, the right intermediate output is the difference between the right input signal and the difference between the left and right input signals modified by the first transfer function, i.e. the right intermediate output is calculated by subtracting the left input signal modified by the first transfer function from the total of the right input signal summed with the right input signal modified by the first transfer function.

[0048] The table below summarises the effect of the processing on the left intermediate signal for the following notional signals when the sound is recorded stereophonically. As before the analysis of the right output follow mutatis mutandis: Signal Condition Transfer function L-*L' 1+H1 R->L' -H1 (L+R)-/L' 1 (L-R)->I2 1 +2H1 As explained above, the most important sounds are usually centrally placed and thus the timbre of centrally placed sounds is unaffected. The increase in the L-R signal leads to a increase in reverberation as highlighted in bold.

[0049] Considering the second stage separately, the output left and right signals L", R" are determined from the intermediate left and right signals L', R': L"=L'-(L'-R')H2=L'-L'H2+R'H2 R"=R'+(12-R)H2=R'-R'H2+L'H2 In this second stage, the summing and differences are reversed when compared to the first stage. Thus, the left output is the difference between the left intermediate signal and the difference between the left and right intermediate signals modified by the second transfer function, i.e. the left output is calculated by subtracting the left intermediate signal modified by the second transfer function from the total of the left intermediate signal summed with the right intermediate signal modified by the second transfer function. Similarly, the right output is the difference between the right input signal and the difference between the left and right input signals modified by the first transfer function, i.e. the right output is calculated by subtracting the right intermediate signal modified by the second transfer function from the total of the right intermediate signal summed with the left intermediate signal modified by the second transfer function.

[0050] The table below summarises the effect of the second stage processing for the left output for the following notional signals when the sound is recorded stereophonically. As before the analysis of the right output follow mutatis mutandis: Signal Condition Transfer function L'->L" 1-H2 R'->L" H2 (12+R)->L" 1 (L'-R')->L" 1-2H2 Thus as highlighted in bold, the cross-feed R'->L" provides lateral localization as desired.

[0051] Considering the two stages together, the output left and right signals L", R" are determined from the input left and right signals L, R (by substituting for L', R' in the equation above), i.e.: L"= L+LH,-RH,-(L+LH,-RH1)H2+(R+RH,-LH1)H2 = L+ LH, -RHO -LH2+ RH2-2LH, H2 + 2RH,H2 = L+ LH1+ RH2+ 2RH,H2-RHO -LH2-2LH,H2 and similarly R"= R+ RHO -LH, -RH2+LH2-2RH, H2 + 2LH1H2 = R + RHO + LH2 + 2LH,H2-LH, -RH2 -2RH,H2 The left output is calculated by subtracting the sum of the right input signal modified by the first transfer function, the left input signal modified by the second transfer function and twice the left input signal modified by both the first and second transfer functions from the total of the left input signal summed with the left input signal modified by the first transfer function, the right input signal modified by the second transfer function and twice the right input signal modified by the first and second transfer functions. Similarly, the right output is calculated by subtracting the sum of the left input signal modified by the first transfer function, the right input signal modified by the second transfer function and twice the right input signal modified by both the first and second transfer functions from the total of the right input signal summed with the right input signal modified by the first transfer function, the left input signal modified by the second transfer function and twice the left input signal modified by the first and second transfer functions. In other words, each output is calculated by subtracting the sum of the opposite side input signal modified by the first transfer function, the same side input signal modified by the second transfer function and twice the same side input signal modified by both the first and second transfer functions from the total of the same side input signal summed with the same side input signal modified by the first transfer function, the opposite input signal modified by the second transfer function and twice the opposite input signal modified by the first and second transfer functions.

[0052] In other words, the left output signal is a weighted sum of the left input signal, each of the right input signal and the left input signal modified by the first transfer function, each of the right input signal and the left input signal modified by the second transfer function and each of the right input signal and the left input signal modified by both the first transfer function and the second transfer function. Similarly, the right output signal is a weighted sum of the right input signal, each of the right input signal and the left input signal modified by the first transfer function, each of the right input signal and the left input signal modified by the second transfer function and each of the right input signal and the left input signal modified by both the first transfer function and the second transfer function. In both cases, the weights are: * 1 for each of the same side input signal, the same side input signal modified by the first transfer function and the opposite side signal modified by the second transfer function, * -1 for each of the same side input signal modified by the second transfer function and the opposite side signal modified by the first transfer function, * 2 for the opposite side input signal modified by both the first and second transfer function and * -2 for the same side input signal modified by both the first and second transfer function.

[0053] The table below summarises the effect of both stages of processing for the left signal the following notional signals when the sound is recorded stereophonically. As before the analysis of the right output follow mutatis mutandis: Signal Condition Transfer function L->L" 1+H1-H2-2H, H2 R->I2 -Hi +112+21H1112 (L+R)->L" 1 (L-R)->L" 1 + 2H1-2H2-41-11I-12 Thus, as intended there is no change to the central sounds because the overall transfer function for (L+R)->L" is unity. By reversing the summing and differences in the second stage when compared to the first stage, the second stage does not simply reverse the effect of the first stage. The presence of the H1H2 which describes an input signal which is modified by both the first and second transfer functions shows that the processing is a sophisticated two stage process.

[0054] Figure 3b illustrates an alternative arrangement which results in the same weighted sum as shown in relation to Figure 3a. In this arrangement, the same two transfer functions H1 and H2 are used but there is no cross-feed between the left and right channels. In other words, the two transfer functions H1 and H2 are applied separately to the left channel and to the right channel and the outputs are then summed to generate the overall left and right outputs L" and R". It will be appreciated that the Figure 3b is a less efficient arrangement to implement but produces the same overall effect. A skilled person may also recognise that there are other alternative arrangements.

[0055] Figure 4 illustrates an alternative second stage of processing which could be used in place of the second stage of Figure 3a. As before, the output left and right signals L", R" are determined from the intermediate left and right signals L', R' but in this arrangement, the overall transfer function for (L+R)->L" for the central sounds is not fixed to unity. This revised arrangement takes into consideration that as shown in Figure la, the ipsilateral and contralateral paths are not identical and thus the central sounds are likely to be affected by this difference in length. Accordingly, a gain setting G is applied to provide the correct timbre at a compromise position between the straight ahead and fully to the side. G is assumed to be less than or equal to unity. It is set to match the extent to which the channels of the source material are correlated. An example method of calculating G is described below in relation to Figures 8a and 8b.

[0056] Considering this new second stage separately, the output left and right signals L", R" are determined from the intermediate left and right signals L', R': L"=L'-(GL'-R')H2=L'-GL'H2+R'H2 R"=R1-(GR'-L')H2=R'-GR'H2+L'H2 Thus, the left output signal is calculated by subtracting the left intermediate signal modified by both the gain setting and the second transfer function from the total of the left intermediate signal summed with the right intermediate signal modified by the second transfer function.

Similarly, the right output signal is calculated by subtracting the right intermediate signal modified by both the gain setting and the second transfer function from the total of the right intermediate signal summed with the left intermediate signal modified by the second transfer function.

[0057] The table below summarises the effect of the second stage processing for the left output for the following notional signals when the sound is recorded stereophonically. As before the analysis of the right output follow mutatis mutandis: Signal Condition Transfer function L'->L" 1-GH2 R'->L" H2 (L'+R')->L" 1+(1-G)H2 (L'-R')->L" 1-(1+G)H2 (L'+GR')->L" 1 Thus as highlighted in bold, the central sounds are now subject to a non-unity transfer function. The transfer function is a function of both the gain setting and the second transfer function. The table includes an additional line to show that the sum of the left intermediate signal and the right intermediate signal to which the gain setting has been applied should result in a unity transfer function. The final line thus represents a typical stereo music signal where the left and right channels are partially correlated. The overall timbre of such a signal would be unaffected by the processing with an appropriate choice for the gain setting. The gain setting may be in the range 0 to 1, more preferably in the range 0.4 to 0.8 (i.e. for correlation of between 70-80%).

[0058] Considering the two stages together, the output left and right signals L", R" are determined from the input left and right signals L, R (by substituting for L', R' in the equation above), i.e.: L"= L+ LI-11+ RH2+ (1+G)RH1H2-RH1 -LGH2-(1+G)LH H2 and similarly R"= R+ RH1+ LH2+ (1+G)LH1H2-LHi -RGH2-(1+G)Rhl1 H2 In other words, each output is calculated by subtracting the combination of the opposite side input signal modified by the first transfer function, the same side input signal modified by the second transfer function and the gain effect, the same side input signal modified by both the first and second transfer functions and the same side input signal modified by both the first and second transfer functions and the gain effect from the total of the same side input signal summed with the same side input signal modified by the first transfer function, the opposite side input signal modified by the second transfer function, the opposite side input signal modified by the first and second transfer functions and the opposite side signal modified by both the first and second transfer functions and the gain effect.

[0059] As before, each right output signal is a weighted combination of the same side input signal, at least one of the right input signal and the left input signal modified by the first transfer function, at least one of the right input signal and the left input signal modified by the second transfer function and at least one of the right input signal and the left input signal modified by both the first transfer function and the second transfer function. In both cases, the weights are: * 1 for each of the same side input signal, the same side input signal modified by the first transfer function and the opposite side signal modified by the second transfer function, * -1 for the opposite side signal modified by the first transfer function, * -G for the same side signal modified by the second transfer function, * 1+G for the opposite side input signal modified by both the first and second transfer function and * -1-G for the same side input signal modified by both the first and second transfer function.

[0060] Figure 5 thus summarises the steps of the processing which can be used in the methods described in Figures 3a and 4. In a first step (S100) a first transfer function, e.g. HI, is selected. The first transfer function may be selected to provide image depth. In a second step (S102) a second transfer function, e.g. H2, is selected. The second transfer function may be selected to provide lateral localisation and may include time delay. In a third optional step (S104), a gain function (G1) may be selected to provide the correct timbre to typical two-channel recordings. The first three steps are shown sequentially but it will be appreciated that they can be done in any order or may be done simultaneously.

[0061] Input left and right signals are received at step S106. The next step is the first stage of the signal processing in which the first transfer function is applied to the input left and right signals to generate intermediate left and right signals (step S108). The intermediate left signal may be a weighted sum of the input left and right signals, with at least one (and maybe both) of the input left signal and the input right signal weighted by the first transfer function.

Similarly, the intermediate right signal may be a weighted sum of the input left and right signals, with at least one of the input left signal and the input right signal weighted by the first transfer function. For example, the intermediate signal may be calculated by subtracting the opposite signal input signal weighted by the first transfer function from the sum of the same side input signal and the same side input signal weighted by the first transfer function.

[0062] The next stage is the second stage of the signal processing in which the second transfer function and any gain effect is applied to the intermediate left and right signals to generate output left and right signals (step S110). The output left signal may be a weighted sum of the intermediate left and right signals, with at least one of the intermediate left signal and the intermediate right signal weighted by the first transfer function. Similarly, the output signal is calculated by subtracting the opposite channel intermediate signal modified by the second transfer function from the total of the same channel intermediate signal summed with the opposite channel intermediate signal modified by the second transfer function.

[0063] Once both stages of processing are completed, the final left and right output signals may be output (step S112). The output left signal may be a weighted sum of the input left and right signals, with at least one (and maybe both) of the input left signal and the input right signal weighted by one or both of the first and second transfer functions (and gain effect where used). Similarly, the output right signal may be a weighted sum of the input left and right signals, with at least one (and maybe both) of the input left signal and the input right signal weighted by one or both of the first and second transfer functions (and gain effect where used). For example, each output signal may be calculated by subtracting the sum of the opposite side input signal weighted by the first transfer function, the same side input signal modified by the second transfer function and twice the same side input signal modified by both the first and second transfer functions from the total of the same side input signal summed with the same side input signal modified by the first transfer function, the opposite input signal modified by the second transfer function and twice the opposite input signal modified by the first and second transfer functions. Where a gain effect is used, the double weighting may be replaced by the sum of 1 and the gain effect. The gain effect may also be applied when applying the second transfer function to the same side input signal.

[0064] Figure 6 is a schematic drawing of a system implementing the method described above. A user 10 has a pair of headphones so that the user's left ear has an adjacent left transducer 12 and the user's right ear has a right transducer 14. The transducers may be any suitable small device which can be worn in or around the user's ears. The transducers may be attached by a band that wraps over or around a user's head. Alternatively, the transducers may be attached only to the user's ears and may thus be termed earbuds or earphones.

Alternatively, the transducers may be integrated into gaming headsets, virtual reality headsets or any other suitable device. The left and right headphones 12, 14 each receive corresponding output left or right channels L" and R" from a signal processing device 20. The signal processing device comprises a processor 20 which may carry out the steps of Figure 5 20 above.

[0065] The signal processing device 20 is shown as a device which is separate from the headphones but it will be appreciated that this is just one arrangement and the components of the signal processing device may be incorporated into the headphones. The signal processing device 20 may comprise standard components such as a memory 24 which may comprise volatile or non-volatile memory and an input/output interface 26. These standard components may be connected by any suitable mechanism such as one or more data buses. The signal processing device 20 may also comprise a data storage 32 which may store software to be implemented on the signal processing device 20. For example, there may be a signal processing module 34 which carries the instructions for the processor to carry out the steps of Figure 5. The input/output interface may output each of the output left and right channel L" and R" from the signal processing device 20 to the appropriate headphone. The input/output interface may also receive the input left and right channels L, R from the sound source 40. The sound source 40 may be any suitable source and may provide a recording or live streamed sound, such as music and/or speech.

[0066] The present invention thus implements the signal processing in software. A hardware solution may also be possible. For example, to get full performance one or more all-pass filters may be used to provide flat delay for H2 over the working frequency range.

[0067] At least some of the example embodiments described herein may be constructed, partially or wholly, using dedicated special-purpose hardware. Terms such as 'component', module', 'processor' or 'unit' used herein may include, but are not limited to, a hardware device, such as circuitry in the form of discrete or integrated components, a Field Programmable Gate Array (FPGA) or Application Specific Integrated Circuit (ASIC), which performs certain tasks or provides the associated functionality. In some embodiments, the described elements may be configured to reside on a tangible, persistent, addressable storage medium and may be configured to execute on one or more processors. These functional elements may in some embodiments include, by way of example, components, such as software components, object-oriented software components, class components and task components, processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuitry, data, databases, data structures, tables, arrays, and variables. Although the example embodiments have been described with reference to the components, modules and units discussed herein, such functional elements may be combined into fewer elements or separated into additional elements.

[0068] As explained above, the second transfer function H2 is selected to improve lateral localization of the recording, e.g. of instruments and voices. As described in "Improved Headphone Listening" by Linkwitz published in Audio, December 1971, one way of achieving a better lateral localization is to design a circuit which reduces the separation between the left and right channels to 3dB for frequencies below 700Hz and then increases above 700Hz. At low frequencies, the wavelengths are long and thus the sound pressure will be in phase at each ear in a natural listening situation because the distance between ears is less than half a wavelength. Nevertheless, there is a difference in intensity in each ear, so channel separation of 3dB is maintained. Figure 7a illustrates such a transfer function and shows that the ipsilateral output (i.e. output to the left headphone transducer from a fully left signal) is a generally flat response with a small increase after 700Hz. By contrast, the contralateral output (i.e. output to the left headphone transducer from a fully right signal) decreases dramatically after 700Hz. The transfer function thus may be considered to be a first order low-pass filter at 700Hz with pass-band gain which is less than 1.

[0069] Figure 7b illustrates the effect of implementing the transfer function shown in Figure 7a as H2. The ipsilateral output and contralateral output are illustrated again. The L+R signal (i.e. the combined output to the left transducer which will be a centrally located instrument or voice) is also shown and as expected from the explanation above, the L+R is a completely flat response. The L-R signal (which is indicative of reverberation) is also shown and as expected the L-R output is considerably reduced at low frequencies. The subjective effect of this is to impair the perception of distance and compress the sound-stage front to back. This is counteracted by the use of a first transfer function HI which is selected to improve depth perception and to improve the depth of the sound-stage front to back.

[0070] Figure 7c illustrates the effect of implementing the transfer function shown in Figure 7a as H1 rather than H2. In this case, the L-R signal is boosted relative to the L+R particularly below 700Hz.

[0071] Figures 8a and 8b illustrate the effect of changing the gain G in the second stage processing shown in Figure 4. The ipsilateral output and contralateral output are illustrated again. As explained above, the cross-feed processing may disturb the frequency response under certain signal conditions. Figure 8a shows the effect of setting G=1 (or replacing the gain stage with a direct connection). The ipsilateral output and contralateral output are unchanged compared to those shown in Figure 7a. The input to the transfer function as shown in Figure 4 is equal to zero for the L+R signal. This ensures a flat response for the L+R signal and this illustrated by the 100% correlated signal in Figure 8a.

[0072] Analysis of typical recordings shows that in fact there is typically only a 70% correlation between the left and right channels. Accordingly, there is unlikely to be the flat response predicted for 100% correlation. This is illustrated in Figure 8a which shows that the 70% correlated signal is generally flat but does have an increase in magnitude above 700Hz. A signal which is not fully correlated may be modelled by using a signal which is 70% L+R and 30% L-R. The output to the left transducer (which is the output shown in Figure 8a) may be modelled by setting the right input signal to have an amplitude of 40% (i.e. (70-30)/(70+30)) of the left input signal.

[0073] Figure 8b illustrates how the 70% correlated signal can be corrected to provide a flat response as desired. In Figure 8b, the value of G has been adjusted to 0.4, G has been calculated using: c -(100 -c) 2c -100 G 100 100 where c is the percentage correlation that we expect in the music recording.

[0074] Various combinations of optional features have been described herein, and it will be appreciated that described features may be combined in any suitable combination. In particular, the features of any one example embodiment may be combined with features of any other embodiment, as appropriate, except where such combinations are mutually exclusive.

Throughout this specification, the term "comprising" or "comprises" means including the component(s) specified but not to the exclusion of the presence of others.

[0075] Attention is directed to all papers and documents which are filed concurrently with or previous to this specification in connection with this application and which are open to public inspection with this specification, and the contents of all such papers and documents are incorporated herein by reference.

[0076] All of the features disclosed in this specification and/or all of the steps of any method or process so disclosed, may be combined in any combination, except combinations where at least some of such features and/or steps are mutually exclusive. Each feature disclosed in this specification may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise. Thus, unless expressly stated otherwise, each feature disclosed is one example only of a generic series of equivalent or similar features.

Claims (23)

18 CLAIMS 1. A method of processing stereo sound for headphones, the method comprising receiving a left input signal and a right input signal, applying a first transfer function to the left and right input signals to generate left and right intermediate signals, applying a second transfer function to the left and right intermediate signals to generate left and right output signals; and outputting the left output signal to a left transducer in the headphones and the right output signal to a right transducer in the headphones; and wherein each of the left and right output signals is generated as a weighted combination of at least one of the right input signal and the left input signal, at least one of the right input signal and the left input signal modified by the first transfer function, at least one of the right input signal and the left input signal modified by the second transfer function and at least one of the right input signal and the left input signal modified by both the first transfer function and the second transfer function.

2. The method of claim 1, wherein the first transfer function is selected to improve a user's perception of depth within the stereo sound.

3. The method of claim 1 or claim 2, wherein the second transfer function is selected to provide lateral localization of individual voices within the stereo sound.

4. The method of any one of the preceding claims, wherein the weighted combination for the left output signal comprises a weight of 1 for each of the left input signal, the left input signal modified by the first transfer function and the right input signal modified by the second transfer function and a weight of -1 for the right input signal modified by the first transfer function.

5. The method of claim 4, wherein the weighted combination for the right output signal comprises a weight of 1 for each of the right input signal, the right input signal modified by the first transfer function and the left input signal modified by the second transfer function and a weight of -1 for the left input signal modified by the first transfer function.

6. The method of claim 4 or claim 5, wherein the weighted combination for the left output signal comprises a weight of -1 for the left input signal modified by the second transfer function, a weight of 2 for the right input signal modified by both the first and second transfer functions and a weight of -2 for the left input signal modified by both the first and second transfer functions.

7. The method of any one of claims 4 to 6, wherein the weighted combination for the right output signal comprises a weight of -1 for the right input signal modified by the second transfer function, a weight of 2 for the left input signal modified by both the first and second transfer functions and a weight of -2 for the right input signal modified by both the first and second transfer functions 8. The method of any one of the preceding claims, further comprising applying a gain effect when applying the second transfer function. 10 9. The method of claim 8, wherein the weighted combination for the left output signal comprises a weight of -G for the left input signal modified by the second transfer function, a weight of 1+G for the right input signal modified by both the first and second transfer functions and a weight of -1-G for the left input signal modified by both the first and second transfer functions 10. The method of claim 8 or claim 9, wherein the weighted combination for the right output signal comprises a weight of -G for the right input signal modified by the second transfer function, a weight of 1+G for the left input signal modified by both the first and second transfer functions and a weight of -1-G for the right input signal modified by both the first and second transfer functions.11. The method of any one of claims 8 to 9 wherein the gain effect is a constant in the range 0.4 to 0.8.12. A non-transitory computer readable medium comprising program instructions which when carried out on a device cause the device to carry out the method of any one of claims 1 to 11.13. A signal processing device comprising: a processor which is configured to receive a left input signal and a right input signal, apply a first transfer function to the left and right input signals to generate left and right intermediate signals, apply a second transfer function to the left and right intermediate signals to generate left and right output signals; and output the left output signal to a left transducer in the headphones and the right output signal to a right transducer in the headphones; wherein each of the left and right output signals is generated as a weighted combination of at least one of the right input signal and the left input signal, at least one of the right input signal and the left input signal modified by the first transfer function, at least one of the right input signal and the left input signal modified by the second transfer function and at least one of the right input signal and the left input signal modified by both the first transfer function and the second transfer function 14. A headphone comprising a signal processing device as claimed in claim 13.Ammendments to the claims have been filed as follows:-CLAIMS1. A method of processing stereo sound for headphones, the method comprising receiving a left input signal and a right input signal, applying a first transfer function to the left and right input signals to generate left and right intermediate signals, applying a second transfer function to the left and right intermediate signals to generate left and right output signals; and outputting the left output signal to a left transducer in the headphones and the right output signal to a right transducer in the headphones; and wherein the left output signal is generated as a weighted combination of the left input signal, the right input signal modified by the first transfer function, the left input signal modified by the first transfer function, the right input signal modified by the second transfer function, the left input signal modified by the second transfer function,the right input signal modified by both the first transfer function and the second transfer function and the left input signal modified by both the first transfer function and the second transfer function, wherein a positive weight is used for each of the left input signal, the left input signal modified by the first transfer function, C\I the right input signal modified by the second transfer function and the right input signal modified by both the first transfer function and the second transfer function and a negative O 20 weight is used for each of the right input signal modified by the first transfer function, the left O input signal modified by the second transfer function and the left input signal modified by both the first transfer function and the second transfer function and the right output signal is generated as a weighted combination of the right input signal, the right input signal modified by the first transfer function, the left input signal modified by the first transfer function, the right input signal modified by the second transfer function, the left input signal modified by the second transfer function, the right input signal modified by both the first transfer function and the second transfer function and the left input signal modified by both the first transfer function and the second transfer function wherein a positive weight is used for each of the right input signal, the right input signal modified by the first transfer function, the left input signal modified by the second transfer function and the left input signal modified by both the first transfer function and the second transfer function and a negative weight is used for each of the left input signal modified by the first transfer function, the right input signal modified by the second transfer function and the right input signal modified by both the first transfer function and the second transfer function.2. The method of claim 1, wherein the first transfer function is selected to improve a user's perception of depth within the stereo sound.3. The method of claim 1 or claim 2, wherein the second transfer function is selected to provide lateral localization of individual voices within the stereo sound.4. The method of any one of the preceding claims, wherein the weighted combination for the left output signal comprises a weight of 1 for each of the left input signal, the left input signal modified by the first transfer function and the right input signal modified by the second transfer function and a weight of -1 for the right input signal modified by the first transfer function.5. The method of claim 4, wherein the weighted combination for the right output signal comprises a weight of 1 for each of the right input signal, the right input signal modified by the first transfer function and the left input signal modified by the second transfer function and a weight of -1 for the left input signal modified by the first transfer function.6. The method of claim 4 or claim 5, wherein the weighted combination for the left output signal comprises a weight of -1 for the left input signal modified by the second transfer function, a weight of 2 for the right input signal modified by both the first and second transfer C\I functions and a weight of -2 for the left input signal modified by both the first and second transfer functions. O 20O 7. The method of any one of claims 4 to 6, wherein the weighted combination for the right output signal comprises a weight of -1 for the right input signal modified by the second transfer function, a weight of 2 for the left input signal modified by both the first and second transfer functions and a weight of -2 for the right input signal modified by both the first and second transfer functions

8. The method of any one of the preceding claims, further comprising applying a gain effect when applying the second transfer function.

9. The method of claim 8, wherein the weighted combination for the left output signal comprises a weight of -G for the left input signal modified by the second transfer function, a weight of 1+G for the right input signal modified by both the first and second transfer functions and a weight of -1-G for the left input signal modified by both the first and second transfer functions

10. The method of claim 8 or claim 9, wherein the weighted combination for the right output signal comprises a weight of -G for the right input signal modified by the second transfer function, a weight of 1+G for the left input signal modified by both the first and second transfer functions and a weight of -1-G for the right input signal modified by both the first and second transfer functions.

11. The method of any one of claims 8 to 10 wherein the gain effect is a constant in the range 0.4 to 0.8.

12. A non-transitory computer readable medium comprising program instructions which when carried out on a device cause the device to carry out the method of any one of claims 1 to 11.

13. A signal processing device for processing stereo sound for headphones comprising: a processor which is configured to receive a left input signal and a right input signal, apply a first transfer function to the left and right input signals to generate left and right intermediate signals, apply a second transfer function to the left and right intermediate signals to generate left and right output signals; and C\I output the left output signal to a left transducer in the headphones and the right output signal to a right transducer in the headphones; O 20 wherein the left output signal is generated as a weighted combination the left input O signal, the right input signal modified by the first transfer function, the left input signal modified by the first transfer function, the right input signal modified by the second transfer function, the left input signal modified by the second transfer function,the right input signal modified by both the first transfer function and the second transfer function and the left input signal modified by both the first transfer function and the second transfer functionwherein a positive weight is used for each of the left input signal, the left input signal modified by the first transfer function, the right input signal modified by the second transfer function and the right input signal modified by both the first transfer function and the second transfer function and a negative weight is used for each of the right input signal modified by the first transfer function, the left input signal modified by the second transfer function and the left input signal modified by both the first transfer function and the second transfer function and the right output signal is generated as a weighted combination of the right input signal, the right input signal modified by the first transfer function, the left input signal modified by the first transfer function, the right input signal modified by the second transfer function, the left input signal modified by the second transfer function, the right input signal modified by both the first transfer function and the second transfer function and the left input signal modified by both the first transfer function and the second transfer function wherein a positive weight is used for each of the right input signal, the right input signal modified by the first transfer function, the left input signal modified by the second transfer function and the left input signal modified by both the first transfer function and the second transfer function and a negative weight is used for each of the left input signal modified by the first transfer function, the right input signal modified by the second transfer function and the right input signal modified by both the first transfer function and the second transfer function.

14. The signal processing device of claim 13, wherein the first transfer function is selected to improve a user's perception of depth within the stereo sound.

15. The signal processing device of claim 13 or claim 14, wherein the second transfer function is selected to provide lateral localization of individual voices within the stereo sound.

16. The signal processing device of any one of claims 13 to 15, wherein the weighted combination for the left output signal comprises a weight of 1 for each of the left input signal, the left input signal modified by the first transfer function and the right input signal modified by the second transfer function and a weight of -1 for the right input signal modified by the first transfer function.OC\I

17. The signal processing device of claim 16, wherein the weighted combination for the right output signal comprises a weight of 1 for each of the right input signal, the right input O 20 signal modified by the first transfer function and the left input signal modified by the second O transfer function and a weight of -1 for the left input signal modified by the first transfer function.

18. The signal processing device of claim 16 or claim 17, wherein the weighted combination for the left output signal comprises a weight of -1 for the left input signal modified by the second transfer function, a weight of 2 for the right input signal modified by both the first and second transfer functions and a weight of -2 for the left input signal modified by both the first and second transfer functions.

19. The signal processing device of any one of claims 16 to 18, wherein the weighted combination for the right output signal comprises a weight of -1 for the right input signal modified by the second transfer function, a weight of 2 for the left input signal modified by both the first and second transfer functions and a weight of -2 for the right input signal modified by both the first and second transfer functions

20. The signal processing device of any one of claims 16 to 19, wherein the processor is configured to further apply a gain effect when applying the second transfer function.

21. The signal processing device of claim 20, wherein the weighted combination for the left output signal comprises a weight of -G for the left input signal modified by the second transfer function, a weight of 1+G for the right input signal modified by both the first and second transfer functions and a weight of -1-G for the left input signal modified by both the first and second transfer functions

22. The signal processing device of claim 20 or claim 21, wherein the weighted combination for the right output signal comprises a weight of -G for the right input signal modified by the second transfer function, a weight of 1+G for the left input signal modified by both the first and second transfer functions and a weight of -1-G for the right input signal modified by both the first and second transfer functions.

23. The signal processing device of any one of claims 20 to 22 wherein the gain effect is a constant in the range 0.4 to 0.8. 15 24. A headphone comprising a signal processing device as claimed in any one of claims 13 to 23. C\ICO O