Development of multichannel single-unit microphone using shotgun microphone array

Transcription

1 PROCEEDINGS of the 22 nd International Congress on Acoustics Electroacoustics and Audio Engineering: Paper ICA Development of multichannel single-unit microphone using shotgun microphone array Yo Sasaki (a), Toshiyuki Nishiguchi (a), Kazuho Ono (a) (a) NHK Science and Technology Research Laboratories, Japan, Abstract We developed a multichannel single-unit microphone using circular shotgun microphone array for simple recording of multichannel audio, such as 22.2ch audio. The microphone array consists of 8 shotgun microphone elements arranged at 45 degree intervals. Though sharp directivity corresponding to the direction of each element is preferable, directivity at low frequency is wide (directivity of a shotgun microphone becomes sharp with increasing frequency due to the acoustic tube). To improve directivity at low frequency, directivity control by digital signal processing can be applied. We evaluated the performance of our multichannel single-unit microphone through numerical simulation and experiment. Keywords: directional control, shotgun microphone, circular microphone array, robustness, regularization

2 Development of multichannel single-unit microphone using shotgun microphone array 1 Introduction NHK is conducting research and development of 8K Super Hi-Vision (SHV) as a nextgeneration medium that offers a highly immersive sense of presence and reality. Test broadcasting of SHV was begun in Japan in 2016 August. The 22.2 multichannel audio (22.2ch audio) is designed for the audio system of SHV and reproduces 3D sound delivering a strong sensation of reality by using a three-layer channel arrangement. The system consists of a top layer with 9 channels, a middle layer with 10 channels, a bottom layer with 3 channels, and 2 channels for low frequency effects (LFEs) [1], as shown in Fig. 1. When recording 22.2ch audio, many microphones are usually used that are separately located to capture sound sources as well as ambient sound. It is, however, not possible to locate many microphones for every situation, such as live recording, where the microphone locations are restricted. It is thus necessary to reduce the space required for the multichannel single-unit microphones as much as possible. To meet this requirement, NHK has developed a portable spherical microphone with acoustic baffles [2] that can record 16ch audio simultaneously. Though it is possible to record multichannel audio corresponding to the direction of sound with the microphone, there is a problem regarding frequency response for the sound arriving from the front due to the effect of acoustic baffles. To solve such a problem, we developed a multichannel single-unit microphone using circular shotgun microphone array. The array consists of 8 shotgun microphones arranged at 45 degree intervals as elements suitable for recording 22.2ch audio. The frequency response is flatter than that of spherical microphones. On the other hand, directivity becomes sharp with increasing frequency, while that at low frequency becomes wide. Directivity control by using Finite Impulse Response (FIR) filter has been applied to improve directivity at low frequency. Directivity improved in the case without any individual difference between the microphone elements. For directivity control, it is preferable to use identical microphone elements with which the controller was designed, but it is not always practical mainly because shotgun microphones are likely to exhibit a difference in sensitivity. To solve this problem, We also propose a robust algorithm for directivity control. 2

3 FIGURE 1. Configuration of 22.2ch audio 2 Multichannel single-unit microphone using circular shotgun microphone array Our multichannel single-unit microphone is shown in Fig. 2. It consists of eight shotgun microphones(sanken CSR-2) as elements, arranged at 45 degree intervals corresponding to the direction of each channel composing 22.2ch audio. The microphone arrangement is suitable for recording ambient sounds, which are likely to be reproduced by 8ch surrounding the listening position per layer, (except FLc and FRc in the mid layer). As shown in Fig. 3, a shotgun microphone has directivity, which becomes shaper as frequency increases due to a leaky acoustic tube. The leaky acoustic tube affects directivity above 1 khz, while directivity below 1 khz is almost the same as that of a cardioid microphone, which is not narrow enough to obtain good separation between channels. Directivity control by using a inverse problem is then applied to improve directivity at low frequency. FIGURE 2. Shotgun microphone array 3

4 FIGURE 3. Frequency response and polar patterns of microphone element (CSR-2) 3 Designing of the Directivity control filter Figure 4 shows a schematic diagram of multichannel single-unit microphone. The outputs of the array are obtained by filtering signals captured by the microphone array where is the number of elements. Suppose a transfer function matrix, which is measured from sound sources located at a sufficient distance to the microphone elements, and target transfer function matrix which are described as (1) (2) where is the transfer function measured from the source located in direction to the element whose main axis is directed to, is the target transfer function from the source located in direction to the microphone element, which is directed to, is the direction of the main axis of the -th elements, is the direction where the -th source is located, and is the angular frequency. Then signals captured by the microphone array are described as, (3) 4

5 where output of the array is described as are the signals of all sound sources. Therefore the (4) where is the transfer function of the directional control filter. On the other hand, the desirable output can be described using as. (5) We calculate optimal that makes the directivity pattern of change to desirable output. If is square and regular, can be described as follows.. (6) If is not square or singular, we calculate minimizing as described below.. (7) FIGURE 4. Block diagram of the Multichannel single-unit microphone 4 Numerical simulation and measurement without sensitivity error We conducted a numerical simulation using measured in a 7.8m 7.6m 6.8m anechoic chamber. A loudspeaker (Genelec 2029A) was located 4 m from the center of the microphone array. Other conditions for designing the filter are as follows. i. The was constructed using transfer functions measured from 360 directions per 5

6 degree. ii. The was constructed using as, (8) iii. iv. where is the amplitude-frequency response of a microphone element for the sound arriving from 0 (front) to maintain the frontal characteristic. The amplitude-frequency response was calibrated by the reference amplitude-frequency response measured using a 1/2 inch free-field measurement microphone (B&K 4191) with a preamplifier (B&K 2670) because the loudspeaker does not have a flat frequency response. Filter length was 4096 taps. The frequency responses for the sound arriving from each direction and polar patterns are plotted in Fig. 5. Directivity narrowed for all directions below 900 Hz. On the other hand, some side-lobes were higher and the directivity patterns became undesirable above 900 Hz. Furthermore, the frequency response for the sound arriving from 0 was not flat. These artifacts are due to spatial aliasing. (b) Polar pattern FIGURE 5. Frequency responses and polar patterns with control over all bands from numerical simulation For avoiding spatial aliasing, it is necessary to limit the frequency range for directivity control. The frequency responses and polar patterns in which the frequency band for directivity control is limited below 900 Hz are plotted in Fig. 6. Although directivity just above 900 Hz was wider than that below 900 Hz, directivity became sharper with increasing frequency due to the effect of a leaky acoustic tube, especially above 4 khz. 6

7 FIGURE 6. Frequency responses and polar patterns with control below 900 Hz from numerical simulation Measurement was conducted under the same conditions as for the numerical simulation above. The frequency responses for the sound arriving from each direction and polar patterns are plotted in Fig. 7. Although the polar pattern at low frequency below about 150 Hz is inferior to that from the simulation, these results show almost good agreement with the simulation. (b) Polar pattern FIGURE 7. Frequency responses and polar patterns with control below 900 Hz from measurement 5 Robustness against individual difference When applying the directivity control at low frequency, it is preferable to use the same microphone elements as those when the controller was designed; however, it is not necessary in practical cases. It is necessary to investigate the effect of the difference in the sensitivity response on the directivity control because shotgun microphones are more likely to exhibit a difference in sensitivity than ordinary microphones due to their structure. We conducted a numerical simulation and measurement in which the microphone elements were arranged in difference order from that used in designing the filter to represent sensitivity difference. The dispersion of the individual sensitivity error each microphone was within ±1 db. 7

8 Figure 8 shows the frequency responses for the sound arriving from each direction and polar patterns with sensitivity error. FIGURE 8. Frequency responses and polar patterns with sensitivity error from measurement The results show that the frequency responses and polar patterns are disturbed at low frequency, especially below 500 Hz. The effect of individual difference increases as frequency decreases. This effect is due to the singularity of due to the thick arrangement of elements compared with the wavelength. It is necessary to stabilize the filter at the frequency band. Therefore, the filter is generally regularized as follows:, (9) where is regularization parameter and is unit matrix. Figure 9 shows the polar patterns and frequency responses with regularization obtained from numerical simulation. Figure10 shows those obtained from measurement. The was determined heuristically. These results show that directivity improves by regularization when there is a sensitivity difference, although the polar patterns become wider compared with those without regularization when there is no sensitivity difference. Regularization is effective in practical cases in which there are likely to be individual differences. 6 Conclusion We proposed a multichannel single-unit microphone using shotgun microphone array. The directivity of a shotgun microphone element becomes sharp as the frequency increases, while it is not sharp enough at low frequency. We applied the directivity control regularized only at low frequency to avoid spatial aliasing. We found that the proposed control algorithm contributes to improving the directivity of practical array microphones even in the case of sensitivity difference. 8

9 FIGURE 9. Frequency responses and polar patterns with regularization from numerical simulation FIGURE 10. Frequency responses and polar patterns with regularization from measurement References [1] K. Hamasaki, T. Nishiguchi, R. Okumura, Y. Nakayama, and A. Ando, "A 22.2 multichannel sound system for ultrahigh-definition TV(UHDTV)," SMPTE J. 117(4), 40-49(2008) [2] K. Ono, T. Nishiguchi, K. Matsui, K. Hamasaki, "Portable spherical microphone for Super Hi-Vision 22.2 multichannel audio,"135th AES convention, Paper Number 8922(2013) 9