x ( Primary Path d( P (z) - e ( y ( Adaptive Filter W (z) y( S (z) Figure 1 Spectrum of motorcycle noise at 40 mph. modeling of the secondary path to

Transcription

1 Active Noise Control for Motorcycle Helmets Kishan P. Raghunathan and Sen M. Kuo Department of Electrical Engineering Northern Illinois University DeKalb, IL, USA Woon S. Gan School of Electrical and Electronic Engineering Nanyang Technological University Singapore Abstract Continuous exposure to high-level of motorcycle noise has a risk of damaging hearing and degrading the quality of music for entertainment. This paper develops an innovative audio-integrated active noise control (ANC) system for motorcycle helmets to cancel noise and preserve the audio quality. Motorcycle noises measured at different speeds are recorded, analyzed and used to evaluate the performance of audio-integrated ANC system. Both computer simulations and real-time experiments show the feasibility of using the developed ANC helmets for hearing protection and audio entertainment. INTRODUCTION According to National Highway Traffic Safety Administration, there are approximately 5 million registered motorcycle owners in USA [1]. Motorcycle noise threatens hearing; it acts as a silent killer and even result in permanent hearing loss. Motorcycle noise levels on an open road are in the ranges of dba at 30 mph (mile-per-hour) to 116 dba at 120 mph, depending on types of helmets, motorcycles, and the conditions of roads. The average rate of increase in noise level is 15.5 db per doubling of speed [2]. Continuous exposure to noise level of 95 dba for more than 2.5 hours may result in various hearing impairments such as, hearing loss, tinnitus, acoustic shock, and even leads to accidents [3]. Figure 1 depicts the spectrum of a real motorcycle noise recorded at the speed of 40 mph, which shows most of energy is concentrated at low frequencies from 10 to 300 Hz. Conventional motorcycle helmets or ear plugs do not provide good attenuation for such low-frequency noise, and it is costly and impractical for certain advanced passive noise attenuation techniques due to weight and size constraints of helmets [4]. In addition, the quality of the audio in the presence of the motorcycle noise degrades drastically, and requires high volume that can even result in mental stress and leads to accident. To attenuate the motorcycle noise and to preserve the quality of audio from the audio system in the motorcycle, this paper presents an innovative audio-integrated ANC technique using feedforward broadband algorithm. Its performance is first evaluated using computer simulations with measured transfer functions from realistic experimental setup, and further tested using realtime experiments using recorded motorcycle noises. Real time experiment of audio-integrated ANC is carried out using recorded real motorcycle sounds as primary noise and music as the desired audio. The objectives of the experimental results are to achieve: 1. Cancellation of harmful noise 2. Integration of audio has no effect on the performance of ANC 3. Desired audio components will not be cancelled by ANC and its quality can be preserved An additional advantage is that the integrated audio masks residual noise after ANC algorithm has converged in steady state. It also provides on-line /10/$ IEEE 170

2 x ( Primary Path d( P (z) - e ( y ( Adaptive Filter W (z) y( S (z) Figure 1 Spectrum of motorcycle noise at 40 mph. modeling of the secondary path to further improve ANC performance under time-varying environment. I. FEEDFORWARD ANC SYSTEM The motorcycle used for acquiring real motorcycle noise is the Harley Davidson s Heritage Softtail with a two-cylinder, 4-stroke engine. The noise was recorded using the 722 high-resolution digital audio recorder [ ]. As shown in Figure 1, motorcycle noise falls into the category of broadband noise. To cancel the broadband noise and to compensate the effects of the secondary path, the broadband feedforward ANC using the filtered-x least-meansquare (FX) algorithm [4] shown in Figure 2 is developed for motorcycle helmets. The undesired noise from a noise source transmits through the primary path P(z) resulting in the primary noise d(, the anti-noise y( generated by the adaptive filter propagates through the secondary path S(z) resulting in y'(. These noises are acoustically superimposed, resulting in error signal (or residual noise) e(, which is fed back to the least-mean-square () algorithm. The reference signal x( is filtered by the estimated secondary path transfer function to generate x'(, which is served as an input to the algorithm. The adaptive filter output is given by T y( w ( x(, (1) w( n ) w0 ( w1 (... w L 1( are the L-tap adaptive filter s coefficients, and where T x ( Figure 2. Block diagram of the broadband ANC using the FX algorithm (adapted from [4]). x( n ) x( x( n 1)... x( n L 1) is the signal vector at time n.the weight-update equation of the adaptive filter is given by w( n 1) w( [ x ( e( ], (2) where is the step size, and the filtered signal vector x '( n ) x'( x'( n 1)... x'( n L 1) T is the filtered version of x (, x ( sˆ( * x(, (3) where s ˆ( is the impulse response of, that is the estimate of the secondary path S(z). To preserve the audio from music players and to mask residual noise after noise cancellation, the traditional ANC algorithm is embedded with audio to form audio-integrated ANC system. There are two important issues to be considered for the integrated system [5]: 1. The audio added to the loudspeaker may interfere with the ANC algorithm 2. ANC system may cancel the desired audio We have to make sure that the audio-integrated ANC system can overcome these problems by removing the audio component from the error signal that is used to update the adaptive filter coefficients. The block diagram of audio-integrated ANC is shown in Figure 3. As shown in Figure 3, the audio component m( is added to adaptive filter output y(. At the acoustic summing junction, the error signal e_m( consists of the residual error e( and the audio component m(. To avoid the interference of the T 171

3 x ( Primary Path d( P (z) e _ m( y ( x ( Adaptive Filter W (z) y( m( Personnel Entertainment system S (z) e ( Figure 3 Block diagram of audio-integrated ANC system. audio component on the ANC algorithm, the audio component m( filtered through secondary path has to be subtracted from e_m( to obtain the true error signal e(. This true error signal is then used to update the weights of the adaptive filter W(z). Assuming that we have the perfect secondary path model, i.e. Sˆ ( z) S( z), the true error becomes e( d( [ s( * y( ]. (4) Thus, the advantages of the audio-integrated ANC system include: (1) eliminating the interference of audio on the performance of traditional ANC, (2) modeling the secondary path online to improve the performance of the FX algorithm, (3) providing audio entertainment and also masking residual noise after noise cancellation, and (4) integrating into the existing helmet ANC hardware such as amplifiers and loudspeakers. This forms a compact, robust, low power, and cost effective solution for ANC in motorcycle helmets and audio on road. II. SIMULATION RESULTS An experimental setup shown in Figure 4 is used to measure the primary path and secondary path transfer functions. The obtained transfer functions are used to construct computer model for simulations. The system consists of one primary microphone, -1 Figure 4 Primary and secondary paths in realistic helmet ANC environment. two error microphones and two secondary speakers as shown (one side only) in Figure 4. The primary speaker is placed 65 cm away from the rider (emulated by B&K torso and head simulator, KEMAR) and the secondary speaker is installed inside the helmet. The primary microphone is placed in front of the helmet near the mouth of the rider, and the error microphone is placed near the ear of the KEMAR. MATLAB simulation of audio-integrated ANC system using the FX algorithm shown in Figure 3 is performed. To prove that the integration of audio does not interfere with the ANC performance, we consider two cases: (1) ANC for motorcycle noise without integration of audio, and (2) ANC for motorcycle noise with integration of audio. The noise to be cancelled is the recorded motorcycle noise, and the music comes from a personnel entertainment system. The filter lengths are 64 taps and 32 taps for the adaptive filter W(z) and the secondary-path estimate, respectively, and the step size is Figure 5 shows spectra of signal measured at the error microphone before (ANC OFF) and after (ANC ON) noise cancellation. It shows that the motorcycle noise has a wide range of frequency components and the average cancellation is approximately 40 db in the frequency range of 0 to 250 Hz. The performance of the system is good in the low-frequency with high levels of the motorcycle noise. 172

4 \ Figure 5 Motorcycle noise spectra at the error microphone without audio integration Figure 7 Power estimation of motorcycle noise at 40 mph, error output and the integrated audio Figure 6 Motorcycle noise spectra at the error microphones with audio integration Figure 6 shows spectra of error signals before and after noise cancellation for the audio-integrated ANC. An important fact observed in Figures 5 and 6 is that similar results for the system with and without audio integration, which proves that the audio integration has no effect on the performance of ANC if the perfect secondary-path model Sˆ ( z) S( z) is available. Figure 7 shows the power estimation of the motorcycle noise at 40 mph, error output, and the integrated audio, which gives the cancellation in terms of power at each instance of time. The convergence of ANC system in terms of power of the error output is observed in Figure 7. The simulation results of audio-integrated feedforward ANC using the FX algorithm are very encouraging. Figure 8 Experimental setup for testing motorcycle helmet ANC III. REAL TIME EXPERIMENTS To mimic a person riding a motorcycle with helmet under noisy conditions, an experimental setup is built as shown in Figure 8. A high-speed 32-bit TMS320C3x signal processor and suitable electronic hardware are selected. B&K s KEMAR is used which resembles a person on a bike wearing a helmet. Primary noise is played through the primary loudspeaker which is placed 65 cm away from the KEAMAR. The error microphone is used to monitor the performance of ANC algorithm. Audio from a music player is directly integrated into the system and is played through the secondary speaker through which anti-noise is played. The real-time experiment results of audiointegrated ANC for real motorcycle noise at 40 mph are shown in Figures 9 and 10. Figure 9 shows the transition from ANC OFF to ANC ON to audio 173

5 ANC OFF ANC ON Audio Integration Figure 9 (i) Spectrogram showing the transition from ANC OFF to ANC ON to audio integration, and (ii) time domain representation of ANC OFF to ANC ON to audio integration. integration using both spectrogram (top) and time waveform. It is observed that the harmful motorcycle noise is cancelled and the audio integration does not interfere with the ANC performance. Two important points observed in Figure 9 are: (1) the integration of audio has no negative effect on the ANC performance, and (2) the spectrogram of audio without noise and spectrogram of background audio after noise cancellation are similar as depicted in Figures 10(i) and 10(iii). This proves that the audio components are not been cancelled by the ANC and the quality of the audio is preserved. Figures 10(iii) shows that the integration of audio with the ANC system not only provide audio entertainment but also masks residual noise after noise cancellation, thus providing pleasant environment for the rider to enjoy music and pay attention on riding. IV. CONCLUSION This paper presented an audio-integrated ANC system to cancel the annoying motorcycle noise and to play the audio such that the audio is not been cancelled by the ANC and the performance of ANC is not degraded. Real motorcycle noises at different speeds were recorded and their characteristics were analyzed. The performance and feasibility of the integrated ANC algorithm was evaluated with the computer simulations using Figure 10 Spectrograms of (i) audio only, (ii) audio with noise (ANC OFF), and (iii) audio with noise (ANC ON). recorded motorcycle noises and measured transfer functions from the experimental setup. Real time experiments of audio-integrated ANC system using the TMS320C3x processor for real motorcycle noise are conducted; the issues and constraints are addressed. Experiment results successfully showed that the integrated audio has no effect on the performance of the ANC and the audio components are not cancelled by the ANC. REFERENCES [1] Highway Statistics [online], U.S. Department of Transportation, Washington, DC. Available: [2] M. C. Lower, D. W. Hurst, A. R. Claughton and A. Thomas, Sources and levels of noise under motorcyclists helmets, Proc. of Institute of Acoustics, vol. 16, part 2, pp , [3] Noise Exposure - Level and Duration [online], The Engineering Toolbox, Available: [4] S. M. Kuo and D. R. Morgan, Introduction to Active Noise Control in Active Noise Control Systems Algorithms and DSP Implementations, ED. New York: Wiley, 1996, pp [5] Lichuan Liu, Shruthi Gujjula, Priya Thanigaim and S. M. Kuo Still in womb: intrauterine acoustic embedded active noise control for infant incubators, Advances in Acoustic and Vibration, Volume 2008(2008), Article ID , March