Audio processing system and method

Abstract

An audio processing system for vehicles generates a frequency domain representation of the audio signal indicative of a first fault condition, having a first edge oriented at a first angle 260, the fr

Description

TECHNICAL FIELD The present disclosure relates to an audio processing system and method. Aspects of the invention relate to audio processing system, a fault identification system, a vehicle, a computer-implemented method, a non-transitory computer-readable medium and computer software. BACKGROUND It is known to process an audio signal originating from a system to identify an audio signature associated with a fault condition. One approach contemplated by the Applicant is to use machine learning algorithms to detect faults from the audio signal in combination with other data collected from one or more vehicle system provided on a vehicle. The machine learning algorithms typically require that the sound data is collected in an appropriate format to enable the accurate analysis to detect a fault condition. In practice, however, the audio signal collected from a vehicle contains background noise and other spurious signals mixed with the sound signatures of the faults. In addition, the sound data might contain signatures from a plurality of fault conditions, such as turbocharger whine, balancer shaft whine, etc. The resulting signal noise may make it difficult to identify a particular fault condition. It is an aim of the present invention to address one or more of the disadvantages associated with the prior art. SUMMARY OF THE INVENTION Aspects and embodiments of the invention provide an audio processing system, a fault identification system, a vehicle, a computer-implemented method, a non-transitory computer-readable medium and computer software as claimed in the appended claims. According to an aspect of the present invention there is provided an audio processing system for processing an audio signal representing soundwaves originating from at least one vehicle system disposed on the vehicle, the audio processing system comprising at least one controller configured to: receive the audio signal; generate a frequency domain representation of the audio signal, the frequency domain representation comprising one or more first feature indicative of a first fault condition, the one or more first feature each having at least one first edge oriented at a first angle; process the frequency domain representation using a first edge detection algorithm associated with the first fault condition, the first edge detection algorithm being direction dependent and configured to detect edges oriented at a first edge detection angle, wherein the first edge detection angle is at least substantially equal to the first angle such that the first edge detection algorithm detects the at least one first edge of the one or more first feature in the frequency domain representation; and generate a first data set representing the at least one first edge detected by the first edge detection algorithm. The audio signal comprises a fault signature associated with the first fault condition. The fault signature may comprise or consist of the one or more first feature indicative of the first fault condition. At least in certain embodiments, the audio processing system may process the audio signal to individually highlight and amplify the one or more first feature. The one or more first feature present in the audio signal may be detected by the first edge detection algorithm. One or more operating parameter of a vehicle system may be used to calculate the first angle of the at least one first edge of the one or more first feature. By determining the angle of the at least one first edge, the edge detection algorithm can isolate and highlight the one or more first feature indicative of the first fault condition. The first data set may comprise or consist of a filtered frequency domain representation in which the or each first edge of the first feature is highlighted. The first data set represents the at least one first edge of the one or more first feature associated with the first fault condition. The identification of the one or more first feature is facilitated by the detection of the at least one first edge. The first data set can be processed to identify the one or more first feature. At least in certain embodiments, the one or more first feature may be identified more readily. This may facilitate the identification of the first fault condition associated with the one or more first feature. The frequency domain representation may comprise a spectrogram. The frequency domain representation may be generated with respect to time. Alternatively, the frequency domain representation may be generated with respect to an operating parameter of one or more of the at least one vehicle system. The first data set may comprise a modified frequency domain representation in which the at least one edge is emphasised to accentuate the one or more first feature. The one or more first feature in the frequency domain representation may be accentuated in the first data set, thereby facilitating identification of the one or more first feature in the frequency domain representation. The first data set may consist of a representation of the at least one first edge detected by the first edge detection algorithm. Data other than the at least one first edge may be excluded from the first data set. Thus, only the at least one first edge may be identified in the first data set. The one or more first feature may be more readily identified in the first data set, thereby facilitating identification of the one or more first feature in the frequency domain representation. The first angle of the first edge may change in dependence on one or more operating parameter of a first vehicle system. The at least one controller may be configured to receive the one or operating parameter of the first vehicle system. The at least one controller may be configured to determine the first edge detection angle in dependence on the one or more operating parameter of the first vehicle system. The first edge detection angle in dependence on a rate of change of the one or more operating parameter of the first vehicle system. The at least one first edge of the one or more first feature may be non-linear in the frequency domain representation. The edge detection algorithm may be dynamic. The first edge detection angle may be determined dynamically in dependence on the one or more operating parameter. The first edge detection angle in dependence on a rate of change of the one or more operating parameter of the first vehicle system. Alternatively, the first angle of the first edge may be constant. The at least one first edge of the one or more first feature may be linear in the frequency domain representation. The one or more first feature comprises or consists of a line or a curve. The one or more first feature may comprise or consist of a fault line or a fault curve. The first edge detection angle may vary in dependence on the associated operating parameter. The analysis may account for these variations when determining the first edge detection angle. The first edge detection angle may be in the range 0° to ±90° inclusive. The first edge detection angle may be substantially zero or may be non-zero. The first edge detection algorithm may apply a kernel comprising a matrix defining a plurality of weights. The kernel may be a direction dependent kernel configured to detect edges oriented at the first edge detection angle. The application of a kernel provides an effective means to detect edges at the first edge detection angle. The edges may be detected using 2 convolution. The first edge detection algorithm may apply the matrix as a mathematical function to the frequency domain representation to determine the first data set. The first data set may express how the characteristics of the frequency domain representation are modified by the matrix. The first data set may comprise a result function representing an integral of the product of the matrix and the frequency domain representation. The kernel may be static (i.e., invariant). The weights defined in the matrix may be fixed or predefined. Alternatively, the kernel may be a dynamically changing kernel. The kernel may change dynamically to change the first edge detection angle. The kernel may be a convolutional kernel. The weights defined in the matrix may vary in dependence on one or more operating parameter of the first vehicle system. The weights defined in the matrix may change dynamically in dependence on changes in the operating parameter of the first vehicle system. The controller may be configured to determine the weights in dependence on a rate of change of the one or more operating parameter of the first vehicle system. The first edge detection angle may change dynamically in dependence on the operating parameter of the first vehicle system. The first edge detection angle may change dynamically in dependence on a rate of change of the operating parameter of the first vehicle system. The operating parameter may comprise an operating speed of the first vehicle system. The first vehicle system may comprise an internal combustion engine. The controller may be configured to determine the weights defined in the matrix in dependence on a rate of change of the operating speed (rpm) of the internal combustion engine. The matrix may comprise or consist of m rows and n columns, wherein m and n are each greater than or equal to ten (10), twenty (20) or thirty (30). By defining the matrix with a greater number of rows and columns, the resolution of the edge detection algorithm may be improved. The accuracy with which the at least one first edge is detected may be improved. The frequency domain representation may comprise one or more second feature indicative of a second fault condition. The one or more second feature may each having at least one second edge oriented at a second angle. The at least one controller may be configured to process the frequency domain representation using a second edge detection algorithm associated with the second fault condition. The second edge detection algorithm may be direction dependent and configured to detect edges oriented at a second edge detection angle. The second edge detection angle may be at least substantially equal to the second angle such that the second edge detection algorithm detects the at least one second edge of the one or more second feature in the frequency domain representation. The second data set may represent the at least one second edge detected by the second edge detection algorithm. The first and second edge detection angles are different from each other. The second angle of the second edge may change in dependence on one or more operating parameter of a second vehicle system. The first and second vehicle systems may be the same as each other or may be different from each other. The at least one controller may be configured to receive the one or operating parameter of the second vehicle system. The at least one controller may be configured to determine the second edge detection angle in dependence on the one or more operating parameter of the second vehicle system. The at least one controller may be configured to determine the second edge detection angle in dependence on a rate of change of the one or more operating parameter of the second vehicle system. The at least one second edge of the second feature may be non-linear in the frequency domain representation. The second edge detection algorithm may be dynamic. The second edge detection angle may be determined dynamically in dependence on the one or more operating parameter of the second vehicle system. The second edge detection angle may be determined in dependence on a rate of change of the one or more operating parameter of the second vehicle system. The at least one controller may be configured to process the frequency domain representation using additional edge detection algorithms, for example associated with a third fault condition. A third edge detection algorithm may be applied to detect edges oriented at a third edge detection angle. The detection of the edges may be at least substantially the same as described herein for the first and second edge detection angles. The audio signal comprises audio data captured by one or more microphone. The one or more microphone may be provided on the vehicle, for example inside a cabin of the vehicle. The audio processing system comprises one or more controllers collectively comprising at least one electronic processor having an electrical input for receiving an input signal; and at least one memory device electrically coupled to the at least one electronic processor and having instructions stored therein; and wherein the at least one electronic processor is configured to access the at least one memory device and execute the instructions thereon so as to: receive the audio signal; generate a frequency domain representation of the audio signal, the frequency domain representation comprising one or more first feature indicative of a first fault condition, the one or more first feature each having at least one first edge oriented at a first angle; process the frequency domain representation using a first edge detection algorithm associated with the first fault condition, the first edge detection algorithm being direction dependent and configured to detect edges oriented at a first edge detection angle, wherein the first edge detection angle is at least substantially equal to the first angle such that the first edge detection algorithm detects the at least one first edge of the one or more first feature in the frequency domain representation; and generate a first data set representing the at least one first edge detected by the first edge detection algorithm. According to a further aspect of the present invention there is provided a fault identification system for identifying a fault condition in a vehicle. The fault identification system may comprise at least one controller configured to receive the first data set from the audio processing system as claimed in any one of the preceding claims. The at least one controller may analyse the first data set to identify a fault condition indicator comprising the at least one first edge detected by the first edge detection algorithm. A fault condition algorithm may be used to analyse the first data set. According to a further aspect of the present invention there is provided a vehicle comprising an audio processing system as described herein. According to a further aspect of the present invention there is provided a computer-implemented method of processing an audio signal representing sound waves originating from at least one vehicle system disposed on the vehicle, the method comprising: receiving the audio signal; generating a frequency domain representation of the audio signal, the frequency domain representation comprising one or more first feature indicative of a first fault condition, the one or more first feature each having at least one first edge oriented at a first angle; processing the frequency domain representation using a first edge detection algorithm associated with the first fault condition, the first edge detection algorithm being direction dependent and configured to detect edges oriented at a 4 first edge detection angle, wherein the first edge detection angle is at least substantially equal to the first angle such that the first edge detection algorithm detects the at least one first edge of the one or more first feature in the frequency domain representation; and generate a first data set representing the at least one first edge detected by the first edge detection algorithm. According to a further aspect of the present invention there is provided a non-transitory computer-readable medium having a set of instructions stored therein which, when executed, cause a processor to perform the method(s) described herein. According to a further aspect of the present invention there is provided computer software that, when executed, is arranged to perform the method(s) described herein. Within the scope of this application it is expressly intended that the various aspects, embodiments, examples and alternatives set out in the preceding paragraphs, in the claims and / or in the following description and drawings, and in particular the individual features thereof, may be taken independently or in any combination. That is, all embodiments and / or features of any embodiment can be combined in any way and / or combination, unless such features are incompatible. The applicant reserves the right to change any originally filed claim or file any new claim accordingly, including the right to amend any originally filed claim to depend from and / or incorporate any feature of any other claim although not originally claimed in that manner. BRIEF DESCRIPTION OF THE DRAWINGS One or more embodiments of the invention will now be described, by way of example only, with reference to the accompanying drawings, in which: Figure 1 shows a schematic representation of a vehicle incorporating an audio processing system in accordance with an embodiment of the present invention; Figure 2 shows a schematic representation of a controller for the audio processing system shown in Figure 1; Figure 3A shows a first spectrogram representing a first audio signal associated with a first fault condition; Figure 3B shows the first spectrogram shown in Figure 3A after being filtered by a first edge detection algorithm in accordance with an embodiment of the present invention; Figure 4A shows a second spectrogram representing a second audio signal associated with a second fault condition; Figure 4B shows the second spectrogram shown in Figure 4A after being filtered by a second edge detection algorithm in accordance with an embodiment of the present invention; Figure 5A shows a third spectrogram representing a third audio signal associated with a third fault condition; Figure 5B shows the third spectrogram shown in Figure 5A after being filtered by a third edge detection algorithm in accordance with an embodiment of the present invention; Figure 6 illustrates the analysis of empirical data to map a fault profile associated with a fault condition; Figure 7 is a schematic representation of the use of convolution to detect an edge; Figure 8 is a schematic representation of a kernel for use in an adaptive convolution process; and Figure 9 shows a first flow diagram representing the operation of the audio processing system to filter a frequency domain representation of an audio signal with respect to one or more operating parameter of a vehicle subsystem. DETAILED DESCRIPTION An audio processing system 1 (not shown) in accordance with an embodiment of the present invention is described herein with reference to the accompanying figures. The audio processing system 1 in the present embodiment is suitable for processing an at least one audio signal AS-n captured by a microphone 105 provided on a vehicle 100. The audio processing system 1 is described herein with reference to the analysis of a first said audio signal AS-1 (not shown). As shown in figure 1, the vehicle 100 in the present embodiment is a road vehicle, such an automobile, a sports utility vehicle or a utility vehicle. The vehicle 100 comprises a plurality of vehicle systems VS-n (not shown). In use, one or more of the vehicle systems VS-n functions as an audio source that emits sound in the form of acoustic waves. The vehicle system(s) VS-n that emit sound waves are referred to herein as vehicle systems VS-n. The sound waves may have frequencies in the audible frequency range (less than approximately 20,000 hertz) and optionally also the ultrasonic frequency range (greater than approximately 20,000 hertz). In use, the microphone 105 captures at least some of the sound waves generated by the vehicle systems VS-n and generates the first audio signal AS-1. The resulting first audio signal AS-1 comprises audio data representing the sound waves emitted by the one or more said vehicle systems VS-n operating on the vehicle 100 at any given time. The microphone 105 in the present embodiment captures the audible sound emitted by the vehicle systems VS-n. In a variant, the microphone 105 could be configured also to capture ultrasonic sound waves for analysis. The audio from the microphone 105 is recorded at its sampling rate. As described herein, the audio processing system 1 is configured to analyse the first audio signal AS-1 to monitor operation of the vehicle systems VS-n. The audio processing system 1 is configured to receive an operating signal OS-n (not shown) indicating an operating parameter of the or each vehicle system VS-n. The audio processing system 1 analyses the first audio signal AS-1 in dependence on the indicated operating parameter of the associated vehicle system VS-n. It will be understood that the audio processing system 1 is operable in conjunction with a range of different vehicle systems VS-n. By way example, the audio processing system 1 according to the present embodiment is described herein with reference to the following: (i) a first said vehicle system VS-1 is in the form of a balancer shaft; (ii) a second said vehicle system VS-2 is in the form of a turbocharger; (iii) a third said vehicle system VS-3 is in the form of a friction brake; and (iv) a fourth said vehicle system VS-4 is in the form of an internal combustion engine. The internal combustion engine VS-4 is provided to generate a propulsive force to propel the vehicle 100. Alternatively, or in addition, the internal combustion engine VS-4 may be provided to charge an onboard traction battery, for example to power a traction battery to propel the vehicle. The balancer shaft VS-1 is an eccentric shaft provided to balance operational loads in the internal combustion engine VS-4. The turbocharger VS-2 is provided to introduce air into the internal combustion engine VS-4 at a pressure greater than atmospheric pressure. Sound waves associated with the operation of each of the first, second, third and fourth vehicle systems VS-1, VS-2, VS-3, VS-4 are detectable in a cabin 111 of the vehicle 100. Other examples of the vehicle system VS-n include an electric traction motor (not shown). For example, the vehicle 100 may be a plug-in hybrid electric vehicle (PHEV) or a battery electric vehicle (BEV) comprising one or more electric traction motor. Other examples of the vehicle system VS-n include a friction brake which, in use, may generate a brake squeal when subject to a fault condition. The processing of the first audio signal AS-n may be performed in dependence on a reference velocity (VREF) of the vehicle 100, for example to account for road noise and / or wind noise detectable in the cabin. In the present embodiment, the microphone 105 is disposed in the cabin 111. The microphone 105 may be a dedicated device for use exclusively with the audio processing system 1. Alternatively, the microphone 105 may be used by one or more other systems, such as an infotainment system. The audio processing system 1 may communicate with a telematic unit on the vehicle 100 to access the audio signal AS-n. By way of example, the microphone 105 may also capture voice commands or audio inputs for a communication system provided on the vehicle 100. It will be understood that the microphone 105 could be provided in other locations of the vehicle 100, for example in an engine bay or an electric traction motor compartment. The audio processing system 1 may receive a plurality of audio signals AS-n, for example from a plurality of the microphones 105 disposed in different locations in the vehicle 100. The audio processing system 1 could be implemented directly on the vehicle 100. For example, one or more controller may be provided on the vehicle 100 to process the audio signal AS-n. Alternatively, the processing of the audio signal AS-n may be performed offboard on a remote server. The data may be output from the vehicle 100 to the remote server for processing. This arrangement may reduce the computational requirements onboard the vehicle 100. The data may be transmitted wirelessly, for example over a wireless communication network; or may be downloaded over a wired connection. The data may be transmitted in real-time. As shown in Figure 2, the vehicle 100 comprises a controller 21 comprising at least one electronic processor 23 and a first system memory 25. The at least one electronic processor 23 has at least one electrical input for receiving vehicle operating signals OS-n (shown in figure 2 as OS-1, OS-2, OS-3 and OS-4) and the audio signal AS-n. The controller 21 is configured to read the vehicle operating signals OS-n from a vehicle communication bus 27, such as Controller Area Network (CAN) bus. The operating signals OS-n comprise operating data indicating a current (i.e., instantaneous) operating parameter of the vehicle systems VS-n (shown in figure 2 as VS-1, VS-2, VS-3 and VS-4). In the present embodiment, a first said operating signal OS-1 indicates an operating speed of the balancer shaft VS-1. A second said operating signal OS-2 indicates a rotational speed of the turbocharger VS-2. A third said operating signal OS-3 indicates a rotational speed of the wheel associated with the friction brake VS-3. A fourth said operating signal OS-4 indicates an operating speed of the internal combustion engine VS-4. The analysis may be performed using one or more of the operating signals OS-1 to OS-4, and one or more additional operating signal OS-n may be captured. The at least one electronic processor 23 is configured to process the first audio signal AS-1. The processing of the first audio signal AS-1 in the present embodiment is performed in dependence on the indicated operating parameter of the associated vehicle system VS-n. The processing of the first audio signal AS-1 may be performed at least substantially in real time. The first audio signal AS-1 generated by the microphone 105 is in a time domain. The at least one electronic processor 23 is configured to transform the audio signal AS-n to a frequency domain. The subsequent analysis of the audio signal AS-n is performed with respect to frequency (rather than time). The frequency domain provides a quantitative indication of the frequency components of the audio signal AS-n. The at least one electronic processor 23 applies a transform, such as a Fourier transform, to decompose the audio signal AS-n into a plurality of frequency components. By way of example, the at least one electronic processor 23 implements a fast Fourier transform algorithm to determine a discrete Fourier transform of the audio signal AS-n. Other transforms may be used to transform the audio signal AS-n. A transform creates a frequency domain representation of the audio signal AS-n. A spectrogram provides a visual 7 representation of the spectrum of frequencies of the audio signal AS-n as it varies with respect to time. The frequency domain representation comprises information about the frequency content of the audio signal AS-n. The magnitude of the frequency components provides an indication of a relative strength of the frequency components. The vehicle system VS-n may develop one or more fault condition. The or each fault condition may change a composition of the sound emitted by the vehicle system VS-n. The fault condition may have an acoustic fault signature which is intrinsically linked to, or associated with, that fault condition. By identifying a particular acoustic fault signature, the associated fault condition can be identified. The acoustic fault signature may comprise or consist of one or more audio component which, either alone or in combination, is indicative of that fault condition. The one or more audio component is captured by the microphone 105 and is present in the audio signal AS-n. By processing the audio signal AS-n, the presence or absence of the one or more audio component can be used to identify the associated fault condition. The identification of the one or more audio component is facilitated by transforming the audio signal AS-n to a frequency domain representation. The frequency domain representation represents the magnitude of the frequency components of the audio signal AS-n and provides an indication of a relative strength of the frequency components. In the present embodiment, the frequency domain representation is generated with respect to time (i.e., a spectrogram). The frequency domain representation is analysed with respect to a known operating parameter of one or more of the vehicle VS-n, such as the operating speed (rpm) of the internal combustion engine VS-4 indicated by the fourth operating signal OS-4. A rate of change of the operating parameter is known at each time in the frequency domain representation. In a variant, the frequency domain representation could be processed to represent the relative strength of the frequency components with respect to an operating parameter of one of the vehicle system VS-n, such as an engine speed (rpm) of the internal combustion engine VS-4. The resulting frequency domain representation may be represented as a function of the oscillation regime, a so-called Campbell or interference diagram. However, this approach requires additional processing which can be avoided by determining the frequency domain representation with respect to time and performing analysis in dependence on a rate of change of the associated operating parameter. The audio processing system 1 is configured to transform the audio signal AS-n into the frequency domain representation. The or each fault condition has an associated (frequency domain) fault profile FP-n (shown in figure 3A as FP-1, figure 4A as FP-2 and figure 5A as FP-3) in the frequency domain representation. The or each fault profile FP-n comprises or consists of a fault line. A first fault profile FP-1 associated with balance shaft whine is shown in a first frequency domain representation in Figure 3A; a second fault profile FP-2 associated with turbocharger whine is shown in a second frequency domain representation in Figure 4A; and a third fault profile FP-3 associated with brake squeal is shown in a third frequency domain representation in Figure 5A. The or each fault profile FP-n in the frequency domain representation may be classified through empirical analysis of one or more audio signals AS-1 captured when a known fault condition is present. The or each fault profile FP-n comprises one or more (frequency domain) feature FDF-n (shown as FDF-1, FDF-2, FDF-2A, FDF-2B and FDF-3). The one or more feature FDF-n is characteristic of a particular fault condition and is mapped to enable identification of the fault condition. By identifying the presence and / or absence of the one or more feature FDF-n in the frequency domain representation, the associated fault profile FP-n can be identified, for example in a diagnostics operation. The or each feature FDF-n comprises an elongated profile, for example in the form of a substantially continuous linear element or curve in the frequency domain representation. The audio processing system 1 is configured to accentuate the or each feature FDF-n in the frequency domain representation to facilitate identification. The one or more feature FDF-n in the frequency domain representation is oriented at a feature orientation angle which is in the range 0° to ±90° (inclusive) relative to a horizontal axis. The feature orientation angle represents a slope or gradient of the or each feature FDF-n. The feature orientation angle is a defining characteristic of the one or more feature FDF-n and enables identification of the one or more feature FDF-n. The feature orientation angle may be determined in 8 dependence on an operating parameter of one or more of the vehicle systems VS-n. For example, the feature orientation angle may be defined with respect to the operating parameter of operating speed (rpm) of the internal combustion engine VS-4. The feature orientation angle of the or each feature FDF-n can be determined with reference to the empirical relationship between the rate of change of the operating parameter and the or each feature FDF-n. For example, it has been determined that the feature FDF-n associated with the balancer shaft VS-1 occurs at a 60th order of the internal combustion engine VS-4 (i.e., a 60th engine order). The slope of the feature FDF-n associated with the balancer shaft VS-1 will occur at approximately sixty (60) times a slope of the operating speed (rpm) of the internal combustion engine VS-4 with respect to time (i.e., a rate of change of the operating speed (rpm)). As described herein, the frequency domain representation is filtered with respect to the known relationship(s) between the rate of change of the operating parameter and the feature orientation angle. In this example, the frequency domain representation may be filtered to highlight one or more feature FDF-n based on the 60th order relationship between the balancer shaft VS-1 and the rate of change of the operating speed (rpm) of the internal combustion engine VS-4. This filter may highlight, or emphasis one or more feature FDF-n associated with a fault condition in the balancer shaft VS-1. The feature orientation angle for the feature FDF-n may vary. In particular, the feature orientation angle may vary over the operating range of the vehicle system VS-n. A feature orientation angle which varies or is non-uniform represents a fault condition in which the frequency components of the audio signal AS-n associated with the fault condition change with respect to the operating parameter of the vehicle system VS-n. A changing or variable feature orientation angle within the frequency domain representation corresponds to a feature FDF-n having a non-linear profile (or a feature FDF-n which may be defined by a non-linear function or approximated by a plurality of linear functions). The feature orientation angle may vary over at least a portion of the operating range of the vehicle system VS-n, for example between upper and lower operating limits. The feature orientation angle varies in dependence on an operating parameter of one or more of the vehicle systems VS-n. By way of example, the feature orientation angle may vary in dependence on a change in an operating speed of one or more of the vehicle system VS-n. The feature orientation angle is defined with respect to a rate of change of the operating parameter of the vehicle system VS-n. The first feature orientation angle in respect of balancer shaft whine (shown in Figure 3A) varies in dependence on a rate of change of an operating speed of the internal combustion engine VS-4. The second feature orientation angle in respect of turbocharger whine (shown in Figure 4A) varies in dependence on a rate of change of an operating speed (rpm) of the turbocharger VS-2 or the internal combustion engine VS-4. The audio processing system 1 in the present embodiment defines the feature orientation angle(s) with respect to the operating parameter of the vehicle systems VS-n, for example in a look-up table or as a mathematical function. Alternatively, the feature orientation angle for a particular feature FDF-n may be fixed. The feature orientation angle may be at least substantially constant across an operating range of the vehicle system VS-n. A constant feature orientation angle represents a fault condition in which the frequency components of the audio signal AS-n associated with the fault condition do not change with respect to the operating parameter of the vehicle system VS-n. A substantially constant feature orientation angle within the frequency domain representation corresponds to a feature FDF-n having a linear profile (or a feature FDF-n which may be defined by a linear function). The feature orientation angle in this example remains substantially constant irrespective of the operating parameter (such as the operating speed) of the vehicle system VS-n over at least a portion of the operating range. An example of a fault profile FP-n comprising a feature FDF-n having a linear profile and a substantially constant feature orientation angle is the fault profile FP-n associated with brake squeal (shown in Figure 5A). The feature orientation angle of the feature FDF-n in the frequency domain representation is zero (0) degrees and the feature FDF-n consists of a substantially horizontal line. It will be understood that other fault conditions may result in one or more feature FDF-n having a feature orientation angle which is substantially constant. A first frequency domain representation FDR-1 of a first audio signal AS-1 is shown in Figure 3A by way of example. The first audio signal AS-1 is associated with a first fault condition in the form of balancer shaft whine emitted from the balancer 9 shaft VS-1. The first audio signal AS-1 is captured over a sample time period. The operating speed (rpm) of the internal combustion engine VS-4 increases smoothly (i.e., increases continuously or progressively) during the sample time period when the first audio signal AS-1 is captured. The operating speed (rpm) of the internal combustion engine VS-4 is recorded with respect to time. A rate of change of the operating speed (rpm) is determined throughout the sample time period. The first frequency domain representation FDR-1 is determined with respect to time. The first frequency domain representation FDR-1 represents the frequency components of the first audio signal AS-1 with respect to time. The first frequency domain representation FDR-1 comprises a first fault profile FP-1 associated with the first fault condition. The first fault profile FP-1 comprises a first feature FDF-1 which is oriented at a first feature orientation angle. The first feature FDF-1 is non-linear and the first feature orientation angle varies in dependence on a rate of change of the operating speed (rpm) of the internal combustion engine VS-4. In a first operating range R1, the first feature FDF-1 is oriented at a first feature orientation angle having a positive value of approximately forty-five degrees (+45°) relative to the horizontal axis. In a second operating range R2, the first feature FDF-1 is oriented at a first feature orientation angle having a negative value of approximately forty-five degrees (-45°) relative to the horizontal axis. The first feature orientation angle is mapped (or defined) with respect to the operating speed (rpm) of the internal combustion engine VS-4. Alternatively, the audio processing system 1 may define the first feature orientation angle with respect to the rotational speed (rpm) of the balancer shaft VS-1. The audio processing system 1 defines the first feature orientation angle with respect to the operating speed (rpm) of the internal combustion engine VS-4. As outlined above, the first feature FDF-1 associated with the balancer shaft VS-1 occurs at a 60th order of the internal combustion engine VS-4 (i.e., a 60th engine order). The slope of the first feature orientation angle is approximately sixty (60) times the slope of the operating speed (rpm) of the internal combustion engine VS-4 with respect to time (i.e., a rate of change of the operating speed (rpm)). The first feature orientation angle can be calculated at any time in the first frequency domain representation FDR-1 in dependence on the known relationship between the rate of change of the operating speed (rpm) of the internal combustion engine VS-4 and the first feature FDF-1. Alternatively, the first feature orientation angle can be defined in a look-up table or by a non-linear function. A second frequency domain representation FDR-2 of a second audio signal AS-2 is shown in Figure 4A by way of example. The second audio signal AS-2 is associated with a second fault condition in the form of turbocharger whine emitted by a turbocharger VS-2. The second audio signal AS-2 is captured over a sample time period. The operating speed (rpm) of the turbocharger VS-2 increases smoothly (i.e., increases continuously or progressively) during the sample time period when the second audio signal AS-2 is captured. The operating speed (rpm) of the turbocharger VS-2 is recorded with respect to time. A rate of change of the operating speed (rpm) is determined throughout the sample time period. The second frequency domain representation FDR-2 is determined with respect to time. The second frequency domain representation FDR-2 represents the frequency components of the second audio signal AS-2 with respect to time. The second frequency domain representation FDR-2 comprises a second fault profile FP-2 associated with the second fault condition. The second fault profile FP-2 comprises a second feature FDF-2 which is oriented at a second feature orientation angle. The second feature FDF-2 is non-linear and the second feature orientation angle varies in dependence on the operating speed (rpm) of the internal combustion engine VS-4. In a first operating range R1, the second feature FDF-2 is oriented at a second feature orientation angle having a positive value of approximately sixty degrees (+60°) relative to the horizontal axis. In a second operating range R2, the second feature FDF-2 is oriented at a second feature orientation angle having a positive value of approximately thirty degrees (+30°) relative to the horizontal axis. The second feature orientation angle is mapped (or defined) with respect to the operating speed (rpm) of the internal combustion engine VS-4. The audio processing system 1 defines the second feature orientation angle with respect to a rate of change of the operating speed (rpm) of the internal combustion engine VS-4. The second feature orientation angle can be calculated at any time in the second frequency domain representation FDR-2 in dependence on a known relationship between the rate of change of the operating speed (rpm) of the internal combustion engine VS-4 and the second feature FDF-2. Alternatively, the second feature orientation angle can be defined in a look-up table or by a non-linear function. It has been recognised that the second fault profile FP-2 associated with the turbocharger whine fault condition may vary in proportion to a number of 10 blades in the turbocharger VS-2. The second feature orientation angle may be determined as a product of the operating speed of the turbocharger VS-2 and the number of blades in the turbocharger VS-. The second feature orientation angle is determined with respect to empirical data. A set of empirical data captured for the second fault condition is represented in Figure 6 by a plurality of discrete points. The second feature FDF-2 comprises a first section FDF-2A and a second section FDF-2B. The second feature FDF-2 is defined as a first best-fit line and a second best-fit line in respect of the first and second sections FDF-2A, FDF-2B respectively. The spectrogram is cropped in the vicinity of the first and second sections FDF-2A, FDF-2B to infer the region in the frequency domain where the second fault condition may occur. A third frequency domain representation FDR-3 of a third audio signal AS-3 is shown in Figure 5A by way of example. The third audio signal AS-3 is associated with a third fault condition in the form of brake squeal generated by the friction brake VS-3. The third audio signal AS-3 is captured over a sample time period. The operating speed (rpm) of the internal combustion engine VS-4 increases smoothly (i.e., increases continuously or progressively) during the sample time period when the third audio signal AS-3 is captured. The operating speed (rpm) of the internal combustion engine VS-4 is recorded with respect to time. A rate of change of the operating speed (rpm) is determined during the sample time period. The third frequency domain representation FDR-3 is determined with respect to time. The third frequency domain representation FDR-3 represents the frequency components of the third audio signal AS-3 with respect to time. The third frequency domain representation FDR-3 comprises a third fault profile FP-3 associated with the third fault condition. The third fault profile FP-3 comprises a third feature FDF-3 in the form of a linear element extending substantially parallel to the horizontal axis (i.e., at a third feature orientation angle of zero (0) degrees). The audio processing system 1 defines the third feature orientation angle with respect to the operating speed (rpm) of the internal combustion engine VS-4. The audio processing system 1 is configured to accentuate, highlight or emphasize the one or more feature FDF-n present in the one or more frequency domain representations FDR-n of the audio signal AS-n. The accentuation of the one or more feature FDF-n may be performed as a pre-processing operation, for example to facilitate the identification of a fault condition in a subsequent processing operation. The or each feature FDF-n comprises at least one edge FDE-n (shown as FDE-1, FDE-2 and FDE-3). The or each edge FDE-n is a discontinuity in the frequency domain representations FDR-n of the audio signal AS-n. In the present embodiment, the audio processing system 1 is configured to implement an edge detection algorithm to detect the one or more edges FDE-n of the or each feature FDF-n. A plurality of the edge detection algorithms may be executed to identify a plurality of the edges FDE-n having different orientations. The operation of the audio processing system 1 to accentuate the one or more feature FDF-n will now be described. The controller 21 is configured to execute one or more edge detection algorithm to identify the at least one edge FDE-n. Each edge detection algorithm is associated with a particular fault condition. For example, a first edge detection algorithm is associated with the first fault condition; a second edge detection algorithm is associated with the second fault condition; a third edge detection algorithm is associated with the third fault condition; and so on. The or each edge detection algorithm is configured to detect an edge oriented at an edge detection angle which corresponds to the feature orientation angle of the associated fault condition. The or each edge detection algorithm is direction dependent. In other words, the or each edge detection algorithm is configured to detect an edge oriented in a predetermined direction. The edge detection algorithm in the present embodiment utilises convolution to detect the one or more edge FDE-n. The edge detection algorithm applies a kernel to the frequency domain representation of the audio signal AS-n. The kernel comprises a matrix defining a plurality of weights which determine an orientation of the edge FDE-n to be detected. The matrix comprises a plurality of m rows and a plurality of n columns. The convolution of an image comprising a horizontal edge EH-1 using convolution is illustrated in Figure 7. A kernel comprising a (3x3) matrix of weights is applied to a source 11 (8x6) image to identify a horizontal edge. The edge detection algorithm outputs a (6x4) data set DS-1 representing the detected edge FDE-1. An enlarged matrix 50 comprising seven (7) rows and seven (7) columns is shown in Figure 8 by way of example. The enlarged (7x7) matrix in this example is configured to detect an edge detection angle of minus half a radian (-0.5 radian). The matrix 50 is defined to ensure that edges can be detected with sufficient angular resolution. A larger matrix 50 can provide improved accuracy. The matrix 50 typically comprises at least ten (10) rows and at least ten (10) columns. In the present embodiment, the matrix comprises fifty-one (51) rows and fifty-one (51) columns. The edge detection algorithm may be a static edge detection algorithm configured to detect an edge having a constant edge detection angle (corresponding to a constant feature orientation angle). The weights defined in the matrix applied by the edge detection algorithm may be static (i.e., fixed) and do not vary with respect to the operating parameter of the one or more vehicle systems VS-n. Alternatively, the edge detection algorithm may be a dynamic edge detection algorithm configured to detect an edge having a variable edge detection angle (corresponding to a variable feature orientation angle). The weights defined in the matrix applied by the edge detection algorithm may be dynamic (i.e., variable) and vary with respect to the operating parameter of the one or more vehicle systems VS-n. The controller 21 is configured to vary the edge detection angle at least substantially to match the feature orientation angle. The controller 21 determines the edge detection angle in dependence on the operating parameter of the one or more vehicle system VS-n, for example in dependence on the operating speed (rpm) of the internal combustion engine VS-4. The controller 21 dynamically adjusts the weights of the matrix 50 to modify the edge detection angle. The edge detection algorithm thereby functions as an adaptive filter which, in the present embodiment, is modified dynamically in dependence on the operating speed (rpm) of the internal combustion engine VS-4. The controller 21 is configured to output one or more data set DS-n representing the or each first edge FDE-1 detected by the first edge detection algorithm. The controller 21 may optionally analyse the or each data set DS-n to identify a fault condition; or the identification of the fault condition may be performed by a separate diagnostics controller. The application of a first edge detection algorithm to accentuate the first feature FDF-1 associated with the first fault condition will now be described by way of example. The first fault condition is in the form of a balancer shaft whine generated by the balancer shaft VS-1. An audio signal AS-n is captured and transformed to a frequency domain representation with respect to the operating speed (rpm) of the internal combustion engine VS-4. The first edge detection algorithm is configured to detect a first edge FDE-1 oriented at a first edge detection angle corresponding to the first feature orientation angle of the first feature FDF-1 in the frequency domain representation. The first edge detection angle is defined to match a predetermined first feature orientation angle for the current (instantaneous) operating parameter of the one or more vehicle system VS-n. The controller 21 in the present embodiment is configured to determine the first edge detection angle in dependence on the operating speed (rpm) of the internal combustion engine VS-4. As outlined above, the relationship between the first edge detection angle and the rate of change of the operating speed (rpm) of the internal combustion engine VS-4 is determined with respect to empirical data. The controller 21 adapts the weights in the kernel in dependence on the operating speed (rpm) of the internal combustion engine VS-4. The first edge detection angle detected by the edge detection angle is thereby modified dynamically with reference to the operating speed (rpm) of the internal combustion engine VS-4. The first edge detection algorithm detects the first edge FDE-1 using convolution, as described herein. The controller 21 is configured to output a first data set DS-1 representing the or each first edge FDE-1 detected by the first edge detection algorithm. The first edge detection algorithm is operative to filter the frequency domain representation of the audio signal AS-n at least partially to remove any components which are not oriented at the first edge detection angle. The first data set DS-1 represents the or each first edge FDE-1 identified in the frequency domain representation of the audio signal AS-n. At least in certain embodiments, the first data set DS-1 may consist of the at least one first edge FDE-1 detected by the first edge detection algorithm. The remainder of the frequency domain representation of the audio 12 signal AS-n may be excluded from the first dataset DS-1. Alternatively, the first edge detection algorithm may be configured to emphasise the at least one first edge FDE-1 relative to the remainder of the frequency domain representation of the audio signal AS-n. The first edge detection algorithm may, for example, highlightthe at least one first edge FDE-1 identified in the frequency domain representation of the audio signal AS-n. The first data set DS-1 comprises a modified representation of the frequency domain representation of the audio signal AS-n in which the or each first edge FDE-1 identified by the first edge detection algorithm is highlighted. The colour or brightness of the or each first edge FDE-1 identified by the first edge detection algorithm may be increased to accentuate the or each first feature FDF-1. Alternatively, or in addition, a contrast between the or each first edge FDE-1 identified by the first edge detection algorithm and the remainder of the frequency domain representation of the audio signal AS-n may be increased. Alternatively, or in addition, a weighting associated with the or each first edge FDE-1 may be increased to facilitate identification. The first edge detection algorithm is operative to pre-process the frequency domain representation of the first audio signal AS-1 to accentuate the or each first feature FDF-1. The operation of the controller 21 to implement the first edge detection algorithm will now be described with reference to a first block diagram 200 shown in Figure 9. The controller 21 receives the operating speed (rpm) of the internal combustion engine VS-4 as a function of time (BLOCK 210). A rate of change of the operating speed (rpm) of the internal combustion engine VS-4 is determined as a function of time (BLOCK 220). The controller 21 determines the relationship between the fault condition and the operating speed (rpm) of the internal combustion engine VS-4 (BLOCK 230). As described herein, the balancer shaft whine may occur at or near a region of 60th to 70th engine order. Balancer shaft whine will be understood to be generated by the balancer shaft / s or the driving mechanism of the balancer shaft / s. The first feature orientation angle is determined as a function of time (BLOCK 240). The first feature orientation angle is calculated by multiplying the rate of change of the operating speed (rpm) of the internal combustion engine VS-4 by the value of the engine order identified through analysis of the empirical data. The controller 21 determines the kernel weights of the matrix 50 as a function of time (BLOCK 250). A spectrogram is generated by transforming the audio signal AS-n from the time domain to the frequency domain. The spectrogram is received by the controller 21 (BLOCK 260). The controller 21 performs adaptive convolution of the spectrogram (BLOCK 270). The adaptive convolution is performed using the dynamically modified kernel weights which are determined as a function of time (BLOCK 250). The controller 21 outputs the filtered spectrogram (BLOCK 280). The filtered spectrogram may, for example, be output as the first data set DS-1. The first data set DS-1 comprises or consists of a filtered frequency domain representation FFDR-1 in which the or each first edge FDE-1 of the first feature FDF-1 is highlighted. The first frequency domain representation FDR-1 of the first audio signal AS-1 associated with balancer shaft whine is shown in Figure 3A. A first filtered frequency domain representation FFDR-1 generated by the application of the first edge detection algorithm by the controller 21 is shown in Figure 3B. The first feature FDF-1 is accentuated in the first filtered frequency domain representation FFDR-1. The first feature FDF-1 can more readily be identified through analysis of the filtered representation of the spectrogram, thereby facilitating the identification of a fault condition associated with balancer shaft whine. The first data set DS-1 comprises or consists of the first filtered frequency domain representation FFDR-1. The second frequency domain representation FDR-2 of the second audio signal AS-2 associated with turbocharger whine is shown in Figure 4A. A second filtered frequency domain representation FFDR-2 generated by the application of the second edge detection algorithm by the controller 21 is shown in Figure 4B. The second feature FDF-2 is accentuated in the second filtered frequency domain representation FFDR-2. The second feature FDF-2 can more readily be identified through analysis of the filtered representation of the spectrogram, thereby facilitating the identification of a fault condition associated with turbocharger whine. The first data set DS-1 comprises or consists of the second filtered frequency domain representation FFDR-2. The third frequency domain representation FDR-3 of the third audio signal AS-3 associated with brake squeal is shown in Figure 5A. A third filtered frequency domain representation FFDR-3 generated by the application of the third edge detection algorithm by the controller 21 is shown in Figure 5B. The third feature FDF-3 is accentuated in the third filtered frequency domain representation FFDR-3. The third feature FDF-3 can more readily be identified through analysis of the filtered representation of the spectrogram, thereby facilitating the identification of a fault condition associated with brake squeal. The first data set DS-1 comprises or consists of the third filtered frequency domain representation FFDR-3. The first data set DS-1 is optionally output to a fault identification system (denoted by the reference numeral 300 in Figure 1). The fault identification system 300 may comprise one or more controller for analysing the first data set DS-1. The fault identification system 300 is configured to process the first data set DS-1 to identify a fault condition indicator comprising one or more of the edge FDE-n identified by the edge detection algorithm(s). It will be appreciated that various changes and modifications can be made to the present invention without departing from the scope of the present application. The audio processing system 1 has been described herein with reference to a controller 21 provided onboard the vehicle 100. It will be understood that the controller 21 may be separate from the vehicle 100. For example, the controller 21 may be incorporated into a vehicle diagnostic system. The audio signal AS-n and / or the first data set DS-1 may be output to the vehicle diagnostic system for processing. The audio signal AS-n may be transformed to the frequency domain by a separate processor. Alternatively, or in addition, the controller 21 may be implemented in a remote server, for example accessed over a suitable communication network.

Claims

1. An audio processing system for processing an audio signal representing sound waves originating from at least one vehicle system disposed on the vehicle, the audio processing system comprising at least one controller configured to: receive the audio signal;generate a frequency domain representation of the audio signal, the frequency domain representation comprising one or more first feature indicative of a first fault condition, the one or more first feature each having at least one first edge oriented at a first angle;process the frequency domain representation using a first edge detection algorithm associated with the first fault condition, the first edge detection algorithm being direction dependent and configured to detect edges oriented at a first edge detection angle, wherein the first edge detection angle is at least substantially equal to the first angle such that the first edge detection algorithm detects the at least one first edge of the one or more first feature in the frequency domain representation; andgenerate a first data set representing the at least one first edge detected by the first edge detection algorithm.

2. An audio processing system as claimed in claim 1, wherein the first data set comprises a modified frequency domain representation in which the at least one edge is emphasised to accentuate the one or more first feature.

3. An audio processing system as claimed in claim 1, wherein the first data set consists of a representation of the at least one first edge detected by the first edge detection algorithm.

4. An audio processing system as claimed in any one of claims 1, 2 or 3, wherein the first angle of the first edge changes in dependence on one or more operating parameter of a first vehicle system, the at least one controller being configured to:receive the one or operating parameter of the first vehicle system; anddetermine the first edge detection angle in dependence on the one or more operating parameter of the first vehicle system.

5. An audio processing system as claimed in any one of the preceding claims, wherein the edge detection algorithm applies a kernel comprising a matrix defining a plurality ofweights, the kernel being a direction dependent kernel configured to detect edges oriented at the first edge detection angle.

6. An audio processing system as claimed in claim 5, wherein the kernel is a dynamically changing kernel.

7. An audio processing system as claimed in claim 6 when dependent directly or indirectly on claim 4, wherein thecontroller is configured to determine the weights in dependence on the operating parameter of the first vehicle system.

8. An audio processing system as claimed in claim 7, wherein the weights change dynamically in dependence on changes in the operating parameter of the first vehicle system.

9. An audio processing system as claimed in claim 8, wherein the operating parameter comprises an operating speed of the first vehicle system.

10. An audio processing system as claimed in any one of claims 5 to 9, wherein the matrix comprises m rows and n columns, wherein m and n are each greater than or equal to ten.

11. An audio processing system as claimed in any one of the preceding claims, wherein the frequency domain representation comprises one or more second feature indicative of a second fault condition, the one or more second feature each having at least one second edge oriented at a second angle, wherein the at least one controller is configured to:process the frequency domain representation using a second edge detection algorithm associated with the second fault condition, the second edge detection algorithm being direction dependent and configured to detect edges oriented at a second edge detection angle, wherein the second edge detection angle is at least substantially equal to the second angle such that the second edge detection algorithm detects the at least one second edge of the one or more second feature in the frequency domain representation; andgenerate a second data set representing the at least one second edge detected by the second edge detection algorithm.

12. A fault identification system for identifying a fault condition in a vehicle, the fault identification system comprising at least one controller configured to:receive the first data set from the audio processing system as claimed in any one of the preceding claims; and analyse the first data set to identify a fault condition indicator comprising the at least one first edge detected by the first edge detection algorithm.

13. A vehicle comprising an audio processing system according to any one of claims 1 to 11.

14. A computer-implemented method of processing an audio signal representing sound waves originating from at least one vehicle system disposed on the vehicle, the method comprising:receiving the audio signal;generating a frequency domain representation of the audio signal, the frequency domain representation comprising one or more first feature indicative of a first fault condition, the one or more first feature each having at least one first edge oriented at a first angle;processing the frequency domain representation using a first edge detection algorithm associated with the first fault condition, the first edge detection algorithm being direction dependent and configured to detect edges oriented at a first edge detection angle, wherein the first edge detection angle is at least substantially equal to the first angle such that the first edge detection algorithm detects the at least one first edge of the one or more first feature in the frequency domain representation; andgenerate a first data set representing the at least one first edge detected by the first edge detection algorithm.

15. A non-transitory computer-readable medium having a set of instructions stored therein which, when executed, cause a processor to perform the method according to claim 14.

16. Computer software that, when executed, is arranged to perform a method according to claim 14.17

Images