Sound design methods for battery-powered electric vehicles, bevs

Abstract

A method, comprising: providing (100) a set of test sound signals (TS) to be played on board a battery-powered electric vehicle, BEV during at least one maneuver thereof, the test sound signals (TS) having a respective set of psychoacoustic characteristics; ranking (110) the set of test sound signals (TS) by assigning a numerical quality score (AR) thereto, and applying regression processing (120) to test sound signals in the set of test sound signals (TS), determining a set of sound parameters (IS, PS, SC, EC) comprising psychoacoustic parameters; wherein a combination of sound parameters in the set of sound parameters (IS, PS, SC, EC) determines the numerical quality score (QS) assigned to the ranked test sounds (AP) in the set of test sounds (TS).

Description

[0001] “Sound design methods for battery-powered electric vehicles, BEVs”

[0002] ****

[0003] TEXT OF DESCRIPTION

[0004] Technical field

[0005] The description relates to methods and systems for sound design based on psychoacoustic parameters, for instance to design sounds for battery-powered electric vehicles, BEVs.

[0006] One or more embodiments may be applied to facilitate satisfying sound quality and / or comfort targets, such as Noise Vibration Harshness, NVH targets according to Society of Automotive Engineers, SAE standards.

[0007] Background

[0008] The lack of the noise of the combustion engine in battery-powered electric vehicles, BEVs, otherwise dominant in many driving conditions, may provide an improved sound comfort to the driver.

[0009] At the same time, people accustomed to driving sporty cars may perceive “silence” as a lack of character. This has led to a shift in car sound design, from noise reduction to how to provide emphasis to car movements. Recent studies have been conducted on the varying influences of sound and vibration on the perceived overall ride comfort, in relation to both electric and combustion engines. Psychoacoustics describes the correlation of the human sensation of sound with its physical sound field parameters. Thereby, pure physical parameters such as level, frequency, bandwidth, duration and degree of modulation are linked to aurally-accurate parameters (psychoacoustic parameters).

[0010] Once the goal of silencing the main sources of noise of a BEV vehicle is achieved, Noise Vibration Harshness (NVH) performance, both in terms of vehicle insulation and sound development, takes on a more prominent role.

[0011] There is a lack of engineering guidelines and directions to perform adequate sound design for reaching acoustic quality and user experience targets.

[0012] Document EP1066623 discusses a process and system for providing objective quality measurement of a target audio signal where reference and target signals are processed by a peripheral ear processor, and compared to provide a basilar degradation signal and in which a cognitive processor employing a neural network determines an objective quality measure from the basilar degradation signal by calculating certain key cognitive model components.

[0013] Document US11718183 discusses a method for controlling a tone of an electric vehicle (EV) based on motor vibration, that may include: calculating an order component from a vibration signal of an EV motor of an electric vehicle, extracting a first order component with the greatest linearity for motor output torque among the calculated order component, then calculating an order frequency by transforming revolutions per minute (RPM) of the EV motor into frequency, setting an EV mode tone by applying a vibration level of the first order component to a level of the order frequency to be output and rearranging the order component, and outputting the set EV mode tone, and may apply an LMS filter algorithm, FFT / IFFT transforms, and an order tracking algorithm in extracting the first order component.

[0014] Existing methods present one or more of the following drawbacks: difficulty in finding a balance between sportiness and comfort; based on subjective liking elements provided by the user, which can hardly be translated into guidelines; difficulty in aligning expectations of a wide public; developments based on time consuming and expensive trial-error approaches.

[0015] Object and summary

[0016] An object of one or more embodiments is to contribute in advancing one or more aspects.

[0017] According to one or more embodiments, that object can be achieved via a method having the features set forth in the claims that follow.

[0018] One or more embodiments may relate to a corresponding system.

[0019] One or more embodiments may include a computer program product loadable in the memory of at least one processing circuit (e.g., a computer) and including software code portions for executing the steps of the method when the product is run on at least one processing circuit.

[0020] As used herein, reference to such a computer program product is understood as being equivalent to reference to a computer-readable medium containing instructions for controlling the processing system in order to co-ordinate implementation of the method according to one or more embodiments.

[0021] Reference to “at least one computer” is intended to highlight the possibility for one or more embodiments to be implemented in modular and / or distributed form.

[0022] The claims are an integral part of the technical teaching provided herein with reference to the embodiments.

[0023] One or more embodiments facilitate the identification of a statistical model and Key Performance Indicators (KPIs) that can serve as technical guidelines for sound designers of BEV vehicles.

[0024] One or more embodiments facilitate to define the main dimensions of the sound that can characterize a BEV sporty vehicle for its users.

[0025] One or more embodiments facilitate performing an analysis of objective data and comparison with the main results of the analysis of subjective data.

[0026] One or more embodiments facilitate to define an automatic tool based on psychoacoustic parameters which aim to describe a BEV sound for sporty vehicles.

[0027] For instance, the tool facilitates computing multilinear correlation between customer appreciation and a deck of selected parameters, mixed in a formula which allow to forecast reaching sound quality targets.

[0028] One or more embodiments facilitate the identification of relevant aspects of the sporty sound and their weighting in contributing to the overall sound quality.

[0029] One or more embodiments facilitates developing a quantitative regressive relation among sound design elements.

[0030] One or more embodiments provide a structured framework to transform inspirations into sound design directions, ultimately enhancing the product delivered to users.

[0031] Brief description of the several views of the drawings

[0032] One or more embodiments will now be described, by way of nonlimiting example only, with reference to the annexed Figures, wherein:

[0033] Figure 1 is a diagram exemplary of a method for identifying audio design knobs according to one or more embodiments;

[0034] Figure 2 is a diagram exemplary of a set of design knobs as per the present disclosure; Figure 3 is a diagram exemplary of an audio quality score as per the present disclosure;

[0035] Figures 4 to 8 are diagrams exemplary of principles underlying one or more embodiments;

[0036] Figure 9 is a diagram exemplary of a sound design method as per the present disclosure; and

[0037] Figure 10, comprising portions a) and b), is a diagram exemplary of spectrograms of sound signals;

[0038] Figure 11 is a diagram exemplary of a battery-powered electric vehicle.

[0039] Corresponding numerals and symbols in the different figures generally refer to corresponding parts unless otherwise indicated.

[0040] The figures are drawn to clearly illustrate the relevant aspects of the embodiments and are not necessarily drawn to scale.

[0041] The edges of features drawn in the figures do not necessarily indicate the termination of the extent of the feature.

[0042] Detailed description

[0043] In the ensuing description, one or more specific details are illustrated, aimed at providing an in-depth understanding of examples of embodiments of this description. The embodiments may be obtained without one or more of the specific details, or with other methods, components, materials, etc. In other cases, known structures, materials, or operations are not illustrated or described in detail so that certain aspects of embodiments will not be obscured.

[0044] Reference to “an embodiment” or “one embodiment” in the framework of the present description is intended to indicate that a particular configuration, structure, or characteristic described in relation to the embodiment is comprised in at least one embodiment. Hence, phrases such as “in an embodiment” or “in one embodiment” that may be present in one or more points of the present description do not necessarily refer to one and the same embodiment.

[0045] Moreover, particular conformations, structures, or characteristics may be combined in any adequate way in one or more embodiments.

[0046] The references used herein are provided merely for convenience and hence do not define the extent of protection or the scope of the embodiments.

[0047] For the sake of simplicity, in the following detailed description a same reference symbol may be used to designate both a node / line in a circuit and a signal which may occur at that node or line.

[0048] As exemplified in Figure 1 , a method comprises: providing 100 a set of test sound signals TS to be played on board a battery-powered electric vehicle, BEV V during at least one maneuver thereof, the test sound signals TS having a respective set of psychoacoustic characteristics; ranking 110 the set of test sound signals TS by assigning a numerical quality score AP thereto, and applying regression processing 120 to test sound signals in the set of test sound signals TS, determining a set of sound parameters IS, PS, SC, EC comprising psychoacoustic parameters; wherein a combination of sound parameters in the set of sound parameters IS, PS, SC, EC determines the numerical quality score QS assigned to the ranked test sounds AP in the set of test sounds TS.

[0049] As exemplified in Figure 1 , applying regression processing comprises providing a set of weights based on a statistical analysis of test sounds in the set of test sound signals and ranking thereof.

[0050] As exemplified herein, the set of weights provided as a result of applying regression processing comprises: a first weight having the greatest value among weight values in the set of weight values; a second weight and a third weight having about the same values, and a fourth weight having the smallest value among weight values in the set of weight values.

[0051] As exemplified herein, sound parameters in the set of sound parameters comprise at least one of: a startup sound parameter contributing with a first weight to the combined quality score, wherein the startup sound parameter is obtained as a combination of the standard deviation of a psychoacoustic articulation index (Al) and a value of the frequency spectrum of the audio signal; a progressive sound parameter contributing with a second weight to the quality score, wherein the progressive sound parameter is a combination of a standard deviation of tonality, of an average value of the sound pressure level; a sportive sound parameter contributing with a third weight to the quality score QS, wherein the sportive sound parameter is a combination of a standard deviation of tonality and at least one average value of the sound pressure level, and an electrical sound parameter contributing with a fourth weight to the quality score, wherein the electrical sound parameter is a linear combination of a root mean square, RMS value of a roughness, a further RMS value of a Tone to Noise value, and a percentage of a prominence value.

[0052] As exemplified herein, ranking the set of test sound signals comprises: receiving a set of digital words comprising a plurality of ranking adjectives associated to test sound signals in the set of test sound signals (TS) via listening tests; applying semantic analysis to digital words in the set of digital words, identifying recurrent digital words as a result, and applying clustering processing to the digital strings of words, assigning adjectives to a set of categories.

[0053] As exemplified herein, applying correlation and variance analysis to rankings of test sounds (TS), preferably comprising R-square and / or adjusted R-square criteria for the analysis.

[0054] As exemplified in Figure 1 , a method for identifying BEV sound design parameters to reach target quality levels comprises: block 100: collecting data indicative of aspects relevant for the overall evaluation of a (e.g., sporty) vehicle’s sound; block 110: performing an analysis of the collected data, determining engineering parameters related to the identified sound aspects, and block 120: providing a set of guidelines of “knobs” to use in sound design, providing multidimensional boundaries within which sound designed are facilitated to operate.

[0055] The method facilitates providing a set of design guidelines to sound engineers, facilitated predicting results of sound quality checks.

[0056] As exemplified in block 100 in Figure 1 , collecting data signals indicative of aspects relevant for the overall evaluation comprises: block 102: holding focus groups comprising pools of experts in order to collect digital strings of words (e.g., adjectives) associated to sound parameters relevant for sport vehicles; block 104: collecting further data related to auditory sensations (e.g., within the same focus groups), and block 106: selecting a set of test sounds TS for further analysis.

[0057] In an exemplary scenario, the test data TS is collected via a set of focus groups (e.g., a preliminary interview with experts) such as three focus groups held with the aim of collecting the main characteristics of a certain type of vehicle (e.g., sporty, premium electric car). For instance, each focus group comprises 6 people with homogeneous sound design expertise within the same company.

[0058] For instance, the operations of collecting 102, 104 and selecting 106 data comprise, for each person in the focus group: identifying a set of natural sounds and associating natural sounds to vehicle types; collecting a set of digital strings of words (e.g., adjectives) deemed to identify the sporty sound from the driver’s perspective; individually providing a set of test sounds TS and associating words in the set of collected words to the test sounds in the set of test sounds TS.

[0059] For instance, the individual operation of playing test sounds TS is performed in a dedicated environment (e.g., a semi-anechoic room) with dedicated equipment (e.g., recorders and technical support for video and sound tracks).

[0060] In an exemplary scenario, the method comprises identifying sound characteristics during different modes of operation (or maneuvers) of a vehicle’s motor (e.g., acceleration, deceleration, constant speed) and / or focusing on designing the sound for the vehicle during a single “reference” mode of operation (e.g., wide open throttle or full throttle).

[0061] An operation of performing an analysis of the collected data TS as exemplified in block 110 of Figure 1 comprises: block 112: ranking of the selected test sounds TS; block 114: performing statistical analysis of the ranking, providing a set of indicators AP that facilitate linking sonority produced and physiology variations in the user (in terms of physical and emotional parameters, using psychoacoustic parameters for instance).

[0062] For instance, the operation of linking acoustic and physiological analysis (that is, psychoacoustic analysis) comprises applying filtering (e.g., low-pass and / or high-pass filtering) to the test sounds and associating the adjectives based on the “frequency signature” or spectrum of test sounds.

[0063] For instance, a jury to perform ranking the test sounds TS exemplified in block 112 comprises a panel of 32 internal experts specializing in Noise, Vibration, and Harshness (NVH) and / or BEV vehicles equipped with headphones and a test grid to fill during the listening.

[0064] For instance, the set of test sounds TS received at block 112 comprises 66 (e.g., randomized) combinations of 12 sounds produced during acceleration of a vehicle.

[0065] For instance, the soundtracks were recorded based on (e.g., five) existing BEV sports cars with sound enhancement and with different sound enhancement modes.

[0066] If possible, the recorded car maneuvers were conducted in both Sound Enhancement OFF and ON conditions. If multiple modes were available, each of them was evaluated.

[0067] For instance, the vehicles (sporty or non-sporty models) may be equipped with sensors for sound recording, such as: binaural headphones worn by the driver; microphones, placed near the driver ears; speed sensors.

[0068] The number of electric motor revolutions was calculated and extracted during post-processing.

[0069] For instance, soundtracks used in one or more embodiments are collected during a run-up car maneuver between 0 and 160 km / h.

[0070] Optionally, it may be possible to collect data related also to nonelectric vehicles and to apply sound post-processing thereto.

[0071] In an exemplary scenario, block 112 further comprises: collecting from jurors also comments for each pair of sounds received for testing; performing semantic analysis of all the strings of comments, identifying recurrent digital words, and applying clustering processing to the digital strings of words, associating comments to a set of categories.

[0072] By combining quantitative statistical data with qualitative insights, sound design guidelines can be drawn to precisely align with the user’s perspective, following a user-centered approach.

[0073] For instance, the following analysis may be performed: analysis of individual comments: evaluating each juror’s specific comments; ranking from test votes: considering the ranking resulting from the test votes, and technical content and modification meaning: analyzing the technical content and the implications of modifications associated with the comments.

[0074] As exemplified in block 120 of Figure 1 , defining a set of objective parameters (or “knobs”) QS to use in sound design (e.g., in order to reach one or more of the defined target sound quality targets) comprises: block 122: for each indicator in the set of indicators AP, performing regression analysis and providing a set of sound design “knobs” parameters SC, IS, PC, EC to match psychoacoustic expectations and sound qualities; block 124: providing a quality score QS based on the values of the set of sound design knobs for the sounds in the test sounds TS.

[0075] In an exemplary scenario, a plurality of aspects is evaluated with an A-B test and ranked using an evaluation (ranging from 0 to 11 ) generated for each subject and sound. For instance, this evaluation represents the number of times each sound was preferred.

[0076] For instance, the regression in block 122 comprises: assessing coherence among data by considering the correlation between each data and the average values of the dataset, excluding outliers as a result; analyzing average ratings and the related dispersion for each of the evaluated aspects, and correlating overall evaluation with objective measures.

[0077] In an exemplary scenario, applying regression in block 122 comprises a Ridge regression method (per se known), which facilitates to mitigate the impact of collinearity.

[0078] Inventors have observed that it may be possible to define a set of objective parameters IS, SC, PC, EC that correlate with the subjective judgments expressed regarding aspects AP during the listening test of test sounds TS.

[0079] In an exemplary scenario, the method 122 comprises applying correlation and variance analysis to rankings of test sounds TS with respect to the aspects AP.

[0080] In an exemplary scenario, block 122 further comprises: performing a correlation analysis among input parameters and a preliminary selection among variables is performed considering only the most important variables from an acoustic point of view, in case two or more parameters are highly correlated; performing a statistical selection of possible models for each of the collected data evaluations (e.g., considering R-square and adjusted R- square criteria); refining the selected models by considering the physical knowledge of the phenomena and psychoacoustic considerations, along with statistical criteria based on model fitting and robustness.

[0081] For instance, the statistical selection one or more of the following considerations may be taken into account: minimum correlation of the parameter with collected data evaluation; variance inflation factor (VI F), significance of each parameter, and coefficient stability with respect to Ridge regressions verify the impact of internal correlations; cross-check by excluding some sounds from the model identification, particularly the influence noises, to assess the impact of outliers and specific sounds.

[0082] As exemplified in Figure 2, a statistical analysis of collected data as exemplified in block 114 may comprise a set of psychoacoustic parameters IS, PC, SC, EC and an associated weight in terms of contribution to the sound quality score QS.

[0083] As exemplified in Figure 2, the set of psychoacoustic parameters IS, PC, SC, EC contributing to the quality score QS comprises: startup or initial sound IS, contributing with a first weight W1 (e.g., 42%) to the quality score QS; progressive feeling PS, contributing with a second weight W2 (e.g., 25%) to the quality score QS; sporting character SC, contributing with a third weight W3 (e.g., 24%) to the quality score QS, and electrical connotation EC, contributing with a fourth weight W4 (e.g., 9%) to the quality score QS.

[0084] As exemplified in Figure 3, the quality score QS can be computed for each of the test sounds and a computed ranking AP’ of the test sounds TS may be reproduced to validate the model developed via the regression analysis 120 (as discussed in the following with reference to Figures 4 to 8).

[0085] As exemplified herein, the initial sound knob IS can be expressed as: IS=a1 *AI_P2dev3 + a2*Max_FFT_3sec + b where a1 , a2, b are coefficients computed as a result of the regression analysis;

[0086] AI_P2dev3 represents the standard deviation of an articulation index Al calculated on a third portion of the sound (starting from its beginning) collected during the recorded vehicle maneuver (e.g., sensed at the driver’s right ear), and

[0087] Max_FFT_3sec is equal to the maximum value of the spectrum of the audio signal calculated (e.g., using Fast Fourier Transform, FFT, known per se) up to the first three seconds of the recorded maneuver sound.

[0088] As appreciable to those of skill in the art, the articulation index AI_P2dev3 is a psychoacoustic parameter.

[0089] Psychoacoustic parameters are known, for instance, from document Huallpa, B., Marano, J., Sczibor, V., de Campos Ferreira, E. et al., “Gear Lever Sound Quality Evaluation,” SAE Technical Paper 2010-36-0369, 2010, doi: 10.4271 / 2010-36-0369.

[0090] For instance, a software tool known under the commercial name of Artemis SUITE and distributed by HEAD acoustics GmbH having headquarters in Herzogenrath, Germany is suitable for computing psychoacoustic parameters.

[0091] For instance, the articulation index AI_P2dev3 is indicative of intelligibility of speech with respect to level and the frequency of background noise.

[0092] As exemplified in portion a) of Figure 4, the contribution of the initial sound knob IS is a linear combination of the sub-knobs Al_p2dev3 and Max_FFT_3sec that contribute in a weighted manner to the overall value based on the coefficients a1 , a2, b extracted via the linear regression processing exemplified in block 124 of Figure 1 .

[0093] As exemplified in portion b) of Figure 4, the observed quality score QS versus predicted values QS’(IS) present a significant alignment.

[0094] In a manner per se known it is possible to compute statistical indicators for the selected knob parameters IS, Al_p3dev3, Max_FFT_3sec yielding a significance level a<0.011 , a coefficient of determination R2=0.63 and a root mean square error (RMSE)=1 .2. The parameters Al_p2dev3 and Max_FFT_3sec have a positive sign, meaning that each of these contributes positively to the global score QS.

[0095] As exemplified herein, the progressive feeling knob PS can be expressed as:

[0096] PS=a1 *L0Uhp_MEDIAmediaT0T + a2*TONIp12_P2devstdTOT + a3*LOU_3_PENDxTEND_P2 + b where

[0097] LOUhp_MEDIAmediaTOT represents an average value of a loudness over time by applying a high pass filter (e.g., cut-off frequency at 1250 Hz) over the entire duration of the maneuver (e.g., average between left and right driver microphone);

[0098] TONIp12_P2devstdTOT represents a standard deviation of a tonality over time by applying a low pass filter (e.g., cut-off frequency 1250 Hz) over the entire duration of the maneuver (e.g., sound collected at driver’s right ear); and

[0099] LOU_3_PENDxTEND_P2 is a loudness level slope (over time) calculated on a third fraction of the total duration of the maneuver (starting from its beginning), multiplied by the first value of the regression line (e.g., for the sound collected at driver’s right ear).

[0100] Figure 5 is a diagram (recorded sound loudness LOU on the ordinate scale, time in seconds in the abscissa scale) having superimposed a computed regression line whose first value R1 is multiplied by the loudness level slope LOU_slope calculated over the third fraction Ttot / 3 of the total time Ttot of the recorded sound in order to obtain the knob parameter LOU_3_PENDxTEND_P2 contributing to the progressive feeling value PS. As appreciable to those of skill in the art, tonality can be determined using Tone to Noise Ratio and Specific Prominence Ratio Analysis use the psychoacoustic concept of frequency groups as hybrids. In one or more embodiments, a loudness method discussed in document ISO, ISO. 532-1 : 2017 “Acoustics — methods for calculating loudness — part 1 : Zwicker method”; 2017 may be used to compute the value of tonality parameter TONIp12_P2devstdTOT. The determination of loudness of stationary signals is performed, for instance, according to the known standard DIN 45631.

[0101] As exemplified in portion a) of Figure 6, the contribution of the progressive feeling knob PS is a linear combination of the sub-knobs LOUhp_MEDIAmediaTOT, TONIp12_P2devstdTOT and a3*LOU_3_PENDxTEND_P2 that contribute in a weighted manner to the overall value based on the coefficients a1 , a2, b extracted via the linear regression processing exemplified in block 120 of Figure 1.

[0102] As exemplified in portion b) of Figure 6, the observed quality score QS versus predicted values QS’(PS) present a significant alignment. In a manner per se known it is possible to compute statistical indicators for the selected knob parameters LOUhp_MEDIAmediaTOT, TONIp12_P2devstdTOT and a3*LOU_3_PENDxTEND_P2 yielding a significance level a<0.002, a coefficient of determination R2=0.84 and a root mean square error (RMSE)=0.84. The knob parameters LOUhp_MEDIAmediaTOT, TONIp12_P2devstdTOT and a3*LOU_3_PENDxTEND_P2 have a positive sign, meaning that each of these contributes positively to the score QS.

[0103] As exemplified herein, the progressive feeling knob PS can be expressed as:

[0104] SC=a1 *TON_MEDIAdevstdTOT+a2*LEhp12_MEDIAmedia1_2tem po+a3*LEV_MEDIAmedia1_3tempo+b where

[0105] TON_MEDIAdevstdTOT represents a standard deviation of tonality calculated against time over the entire duration of the maneuver (e.g., as an average between left and right driver microphone);

[0106] LEhp12_MEDIAmedia1_2tempo represents the average value of the sound pressure level calculated against time (e.g., by applying a high pass filter 1250 Hz) on the first half of the maneuver (e.g., as an average between left and right driver microphone);

[0107] LEV_MEDIAmedia1_3tempo represents an average of the sound pressure level calculated against time on the first third of the maneuver (average between left and right driver microphone).

[0108] As exemplified in portion a) of Figure 7, the contribution of the sportive character knob SC is a linear combination of the sub-knobs TON_MEDIAdevstdTOT, a2*LEhp12_MEDIAmedia1_2tempo, a3*LEV_MEDIAmedia1_3tempo that contribute in a weighted manner to the overall value based on the coefficients a1 , a2, b extracted via the linear regression processing exemplified in block 124 of Figure 1.

[0109] As exemplified in portion b) of Figure 7, the observed quality score QS versus predicted values QS’(PS) present a significant alignment.

[0110] In a manner per se known it is possible to compute statistical indicators for the selected knob parameters TON_MEDIAdevstdTOT, a2*LEhp12_MEDIAmedia1_2tempo, a3*LEV_MEDIAmedia1_3tempo yielding a significance level a<0.001 , a coefficient of determination R2=0.89 and a root mean square error (RMSE)=0.79. The knob parameters LOUhp_MEDIAmediaTOT, TONIp12_P2devstdTOT and a3*LOU_3_PENDxTEND_P2 have a positive sign, meaning that each of these contributes positively to the score QS.

[0111] As exemplified herein, the electrical connotation knob EC can be expressed as:

[0112] EC=a1 *ROU_RMS_P2+a2*TtN_RMS_0_5000Hz_P2+a3*PRO_SU P_10_P2_pct+b where

[0113] ROU_RMS_P2 represents a root mean square (RMS) value of a roughness calculated over time for the entire duration of the maneuver (e.g. , sensed at driver’s right ear);

[0114] TtN_RMS_0_5000Hz_P2 represents a RMS value in a certain frequency band (e.g., from 0 to 5000 Hz) of a Tone to Noise value calculated over time for the entire duration of the maneuver (e.g., sensed at driver’s right ear); and

[0115] PRO_SUP_10_P2_pct represents a percentage of a prominence / impulsiveness value exceeding 10 dB in a certain frequency band (e.g., between 100 and 5000 Hz of the sound collected at driver’s right ear).

[0116] As appreciable to those of skill in the art, the parameter of impulsiveness is generated by fast and significative signal level fluctuations in the audio. Therefore, the psychoacoustic quantity Impulsiveness maps the human sensation of fast and at the same time huge noise level changes to a linear scale.

[0117] As appreciable to those of skill in the art, roughness is a psychoacoustic quantity that maps the human physiological perception of abnormalities of acoustical signals to a linear scale. For instance, impulsiveness increases: with increasing signal level, with increasing relative impulse exaggeration, which may be expressed as a / (1 - a); with increasing relative impulse width p, reaches a maximum and decreases again, and for impulse functions with a steeper slope (e.g., rectangle oscillations).

[0118] As exemplified in portion a) of Figure 8, the contribution of the electrical connotation / character knob EC is a linear combination of the subknobs ROU_RMS_P2, TtN_RMS_0_5000Hz_P2, PRO_SUP_10_P2_pct that contribute in a weighted manner to the overall value based on the coefficients a1 , a2, b extracted via the linear regression processing exemplified in block 124 of Figure 1.

[0119] As exemplified in portion b) of Figure 8, the observed quality score QS versus predicted values QS’(ES) present a significant alignment.

[0120] In a manner per se known it is possible to compute statistical indicators for the selected knob parameters ROU_RMS_P2, TtN_RMS_0_5000Hz_P2, PRO_SUP_10_P2_pct yielding a significance level a<0.001 , a coefficient of determination R2=0.95 and a root mean square error (RMSE)=0.39. The knob parameters LOUhp_MEDIAmediaTOT, TONIp12_P2devstdTOT and a3*LOU_3_PENDxTEND_P2 have a positive sign, meaning that each of these contributes positively to the score QS.

[0121] In one or more embodiments, the (coefficients for each of the) objective parameters PS, SC, EC, IS can be obtained via an artificial neural network, ANN stage configured to apply regression processing to sound signals and / or pattern recognition to sound maps. For instance, the ANN stage may be trained to perform regression / pattern recognition with the data TS, AP, QS collected as exemplified in Figure 1.

[0122] Inventors have observed that the beginning of each car maneuver reveals a delicate aspect, as even slight changes in level trends can alter subjective appreciation.

[0123] Inventors have further observed that high frequency can affect the judgment at the start of the maneuver while the sudden growth in level can be appreciated in the judgment at the start.

[0124] For instance, a method 90 as exemplified in Figure 9 comprises: providing at least one sound signal ST to be played on board a battery-powered electric vehicle V during at least one maneuver thereof, the at least sound signal having a set of psychoacoustic characteristics; applying signal processing 91 , 92, 93, 94 to the sound signal, measuring a set of sound parameters SC, IS, PC, EC comprising psychoacoustic characteristics as a result; based on the set of sound parameters, computing 95 a combined psychoacoustic quality score QS of the sound signal; performing a comparison 96 of the quality score and at least one reference value QS_target, and varying 97, 98 the set of psychoacoustic characteristics of the sound signal based on the result of said comparison.

[0125] As exemplified in Figure 9, a method for semi-automated sound design based on the knobs IS, SC, PC, EC identified with the method exemplified in Figure 1 , comprises: receiving a soundtrack ST for reproduction on-board a BEV; applying (e.g., digital) signal processing 90 to the received soundtrack ST, determining whether the soundtrack is acceptable or not based on a computed quality score QS, and modifying 98 the soundtrack ST based on a difference between the computed quality score QS and a target quality score QS_target, wherein the parameters IS, SC, PC, EC of the soundtrack ST that are modified are those that present a lower value in contributing to the quality score QS.

[0126] As exemplified in Figure 9, the signal processing tool 90 comprises: a set of signal processing stages 91 , 92, 93 configured to extract from the sound the values of objective or knob parameters IS, PS, SC, EC; a weighted addition stage 95 configured to provide the sound quality score QS based on the weighted sum of the knob parameters IS, PS, SC, EC where each knob parameter in the set of knob parameters is multiplied by a respective weight factor W1 , W2, W3, W4 before computing a global sum; a comparison stage 96 configured to perform a comparison between the computed sound quality score QS and a target sound quality level QS_target (e.g., 6 or higher), for instance the comparison stage 96 being configured to compute a difference A=QS-QStarget; a checkout stage 97 configured to provide an indication that the soundtrack ST is ready to be equipped on board the BEV in case the difference A is zero or below a given value (e.g., 2) and / or configured to send the values of the computed knob parameters IS, EC, PC, SC to a user sound design tool 98.

[0127] As exemplified in Figure 9, the user sound design tool 98 can be configured to vary the soundtrack parameters based on the received set of knob parameters IS, SC, PC, EC providing guidance to the sound design engineer to increase those knob values that are lower or that are weighted higher (e.g., first knob IS may be the lowest and weighted 40%) to arrive at the target quality score QS_target.

[0128] As exemplified in Figure 9, the sound design tools 90, 98 can be automated tools implemented on a computer.

[0129] Portion a) of Figure 10 represents a spectrogram of an initial sound ST provided by the sound designer to the tool 90. As appreciable by a person skilled in the art, a spectrogram is represented as a heat map (i.e., as an image with the intensity shown by varying the colour or brightness) in which the ordinate scale represents the speed of the vehicle VM(t) over time (in meters per second), and the abscissa scale represents frequency (in Hertz); a third dimension indicating the amplitude of a particular frequency at a particular time is represented by the intensity or color of each point in the image. The frequency and amplitude axes can be either linear or logarithmic. For instance, the spectrograms exemplified in Figure 10 can be create using the FFT via a digital process, in a manner per se known. As exemplified in portion a) of Figure 1 , speed-related noise shows up as sloped lines in the plot. For instance, for a four-cylinder, four-stroke engine, the dominant frequency is twice the speed of rotation or the second harmonic. In the case of rotating machinery, this can be called “second order”.

[0130] As a result of analysis via the tool 90 exemplified in Figure 9, the following objective parameter values can be obtained for the sound ST exemplified in portion a) of Figure 10:

[0131] TON_MEDIAdevstdTOT(ST)=0.60

[0132] LEhp12_MEDIAmedia1_2tempo=40.06

[0133] LEV_MEDIAmedia1_3tempo=61 .48

[0134] As a result of the analysis, the score QS has a first value (e.g., QS(ST)=4) which is clearly influenced by the first value, which is the lowest and contributes to the sporting character sound parameter SC. Therefore, the tool 90 provides to the sound designer an indicator of how to improve the acoustic quality of the sound to satisfy the quality target score QS_target in terms of psychoacoustic parameters.

[0135] Portion b) of Figure 10 represents a further spectrogram of a sound ST1 in which the value of the first objective parameter “TON_MEDIAdevstdTOT” has been varied to score a higher quality score. The variation in the sound can be appreciated by visually comparing the slopes of the left parts of the spectrograms a) and b) of Figure 10.

[0136] As a result of further refining, the following objective parameter values can be obtained for the modified sound ST1 exemplified in portion b) of Figure 10:

[0137] TON_MEDIAdevstdTOT(ST)=1 .01

[0138] LEhp12_MEDIAmedia1_2tempo=39.93

[0139] LEV_MEDIAmedia1_3tempo=64.56

[0140] Therefore, the quality score for the modified sound ST 1 has a second value (e.g., QS(ST1 )=6) closer to the target score QS_target (e.g., QS_target=8) with respect to the original sound ST (e.g., QS(ST)=4).

[0141] As exemplified in Figure 11 , a battery-powered electric vehicle V comprises an electric motor M and a control unit ECU for driving at least one loudspeaker LS to play (e.g., internally for the driver and / or externally for the pedestrians) the sound ST1 obtained via the method / tool 90 exemplified in Figure 9.

[0142] It will be otherwise understood that the various individual implementing options exemplified throughout the figures accompanying this description are not necessarily intended to be adopted in the same combinations exemplified in the figures. One or more embodiments may thus adopt these (otherwise non-mandatory) options individually and / or in different combinations with respect to the combination exemplified in the accompanying figures.

[0143] Without prejudice to the underlying principles, the details and embodiments may vary, even significantly, with respect to what has been described by way of example only, without departing from the extent of protection. The extent of protection is defined by the annexed claims.

Claims

CLAIMS1. A method, comprising: providing (100) a set of test sound signals (TS) to be played on board a battery-powered electric vehicle, BEV (V) during at least one maneuver thereof, the test sound signals (TS) having a respective set of psychoacoustic characteristics; ranking (110) the set of test sound signals (TS) by assigning a numerical quality score (AP) thereto, and applying regression processing (120) to test sound signals in the set of test sound signals (TS), determining a set of sound parameters (IS, PS, SC, EC) comprising psychoacoustic parameters; wherein a combination of sound parameters in the set of sound parameters (IS, PS, SC, EC) determines the numerical quality score (QS) assigned to the ranked test sounds (AP) in the set of test sounds (TS).

2. The method of claim 1 or claim 2, wherein applying regression processing (122) comprises applying a Ridge regression method.

3. The method of claim 1 , wherein applying regression processing (120) comprises providing (122, 124) a set of weights (W1 , W2, W3, W4) based on a statistical analysis (110, 120) of test sounds in the set of test sound signals (TS) and ranking thereof.

4. The method of claim 3, wherein the set of weights (W1 , W2, W3, W4) provided as a result of applying regression processing (122, 124) comprises: a first weight (W1 ) having the greatest value among weight values in the set of weight values (W1 , W2, W3, W4); a second weight (W2) and a third weight (W3) having about the same values, and a fourth weight (W4) having the smallest value among weight values in the set of weight values (W1 , W2, W3, W4).

5. The method of claim 4, wherein the first weight value (W1 ) is about 42%; the second weight value (W2) is about 25%; the third weight value (W3) is about 24%, and the fourth weight value (W4) is about 9%.

6. The method of any one of the previous claims, wherein the at least one maneuver of the vehicle (V) comprises a wide-open throttle accelerationmaneuver, preferably collected during a run-up car maneuver between 0 and 160 km / h.

7. The method of claim 1 , wherein sound parameters in the set of sound parameters (SC, IS, PC, EC) comprise at least one of: a startup sound parameter (IS) contributing with a first weight (W1 ) to the combined quality score (QS), wherein the startup sound parameter (IS) is obtained as a combination of the standard deviation (AI_P2dev3) of a psychoacoustic articulation index (Al) and a value of the frequency spectrum (Max_FFT_3sec) of the audio signal; a progressive sound parameter (PS) contributing with a second weight (W2) to the quality score (QS), wherein the progressive sound parameter (PS) is a combination of a standard deviation of tonality (TON_MEDIAdevstdTOT), of an average value of the sound pressure level; a sportive sound parameter (SC) contributing with a third weight (W3) to the quality score QS, wherein the sportive sound parameter (SC) is a combination of a standard deviation of tonality (TON_MEDIAdevstdTOT) and at least one average value of the sound pressure level (LEhp12_MEDIAmedia1_2tempo, LEV_MEDIAmedia1_3tempo), and an electrical sound parameter (EC) contributing with a fourth weight (W4) to the quality score (QS), wherein the electrical sound parameter (EC) is a linear combination of a root mean square, RMS value of a roughness (ROU_RMS_P2), a further RMS value of a Tone to Noise value (TtN_RMS_0_5000Hz_P2), and a percentage of a prominence value (PRQ_SUP_10_P2_pct).

8. The method of any one of the previous claims, wherein ranking (110) the set of test sound signals (TS) comprises: receiving a set of digital words comprising a plurality of ranking adjectives associated to test sound signals in the set of test sound signals (TS) via listening tests; applying semantic analysis to digital words in the set of digital words, identifying recurrent digital words as a result, and applying clustering processing to the digital strings of words, assigning adjectives to a set of categories, preferably wherein the method further comprises applying filtering to the test sounds and associating ranking adjectives based on a spectrum oftest sounds.

9. The method of any one of the previous claims, comprising applying correlation and variance analysis to rankings of test sounds (TS), preferably comprising R-square and / or adjusted R-square criteria for the analysis.

10. A computer-readable medium comprising instructions which, when executed by a processing device, cause the processing device to carry out the operations of the method according to any one of claims 1 to 9.

Images