Contributions of various transmission paths to speech privacy of open ceiling meeting rooms in open-plan offices

Installing open ceiling meeting rooms inside a large open-plan office provides a solution to increase speech privacy and to reduce speech disturbance in the office. The open ceiling meeting rooms have advantages of low cost construction and flexibility, but have lower speech privacy than that of enclosed rooms due to the open ceiling. Existing research shows that many factors should be taken into account to achieve good speech privacy in open-plan offices and improving only one of these factors may result in little improvement, so it is important to distinguish contributions of different acoustic transmission paths of open ceiling meeting rooms in open-plan offices. This paper proposes an impulse response separation method to quantify contributions of various acoustic paths of open ceiling rooms on speech privacy in open-plan offices. The method is verified with simulations based on the Odeon software and the experiments carried out in 3 different types of rooms. Finally, the proposed method is applied to the Fabpod, a semi enclosed meeting room located in a large indoor office at the Design Research Institute of the RMIT University, to obtain the contributions of different acoustic transmission paths to its speech privacy. The method proposed in this paper and the knowledge obtained are useful for architects to improve the acoustic performance of the next generation Fabpods which are now under design at RMIT University.


Introduction
Since late 1960s, open-plan offices have been popular among design professionals [1]. Large open-plan offices have advantages of low cost construction and flexibility, but sometimes they lack speech privacy and result in speech disturbance when people are talking. Installing small closed meeting rooms inside open-plan offices provides a solution to the problem; however, the ceiling increases the cost of the meeting rooms due to the requirements of fire safety regulations and extra ventilation and lighting systems.
Keeping the ceiling open or removing the ceiling of meeting rooms is an option; but the challenge is the low speech privacy due to sound propagating out through the open ceiling. There are several acoustic transmission paths through which sound radiates out from open ceiling meeting rooms into open-plan offices, and their relative contributions to speech privacy will be analyzed in this paper.
Speech privacy is related to the speech to noise ratio and represents the opposite of Speech Intelligibility (SI) to some extent [2]. In North America, Articulation Index (AI) and the Speech Intelligibility Index (SII) are widely used to represent the speech privacy while the Speech Transmission Index (STI) is used in Europe to represent speech privacy in open-plan offices [3]. STI is a physical quantity representing the transmission quality of speech with respect to intelligibility, and this paper uses it to evaluate the speech privacy of open ceiling meeting rooms in open-plan offices [4].
The relationship between room acoustic parameters and speech privacy of open-plan offices has been investigated by some researchers [5][6][7]. An international measurement standard was published in 2012, which uses single number quantities to indicate the general acoustic performance of open-plan offices [5]. The converted four single number quantities are the distraction distance, the spatial decay rate of speech, the A-weighted Sound Pressure Level (SPL) of speech at 4-m-distance and the average A-weighted background noise level, and can be determined by the spatial curves of A-weighted SPL of speech and STI in the office [6]. On the other hand, these single number quantities can be estimated by physical and acoustic parameters of rooms, which include the length, width, height of the room, the ceiling absorption, the screen height and apparent furnishing absorption [7]. 3 To achieve good speech privacy performance, many room acoustic parameters should be considered at the same time and improving only one of these factors may result in little improvement if it is not the most critical one [3]. To identify the most critical factor, it is necessary to explore the influence of each parameter separately. Acoustical elements that can affect the acoustical environment in open-plan offices, such as windows, walls, ceilings and partial height screens, have been investigated experimentally [8]. But these experimental case studies lack quantitative analysis, which makes it hard to consider all important factors at the same time and compare the influence of different room acoustic parameters. An alternative way is to develop analytical models. A simple model of a single screen in an open-plan office with ceiling and floor reflections has been developed by using the image source technique [9]. A more complicated model took the effects of side and back panels of the common separating screen into account, and was used to investigate the sound propagation between two adjacent rectangular workstations in an open-plan office [10]. Some models even considered wall reflections and reverberation [11].
There are many acoustic transmission paths for open ceiling meeting rooms to radiate sound out into open-plan offices. The paths of reflecting from the ceiling and diffracting over the panel are relatively important while transmitting through the panel, reverberating in the room and reflecting by office equipment cannot be ignored either [11]. Based on the analytical models, the ceiling sound absorption, the panel height of the open-plan office and the office size were found to be the most important factors, while panel absorption, panel transmission loss, floor absorption, ceiling height and the details of ceiling mounted lighting could not be ignored though less important [2]. By optimizing all these room acoustic parameters simultaneously, good acoustic design can be obtained to meet the criterion for acceptable speech privacy.
The acoustic impulse responses of a room can provide most important acoustic information of the room [12]. For example, some important room acoustic parameters like reverberation time can be estimated from the room impulse responses [13].
Commercial room acoustic software such as Odeon and Dirac can be used to obtain a variety of parameters from the impulse responses [14,15]. Bradley used the impulse responses to describe energy diffracted by the panels and reflected by the ceiling to 4 compare their influence on speech privacy in actual rooms [3]. But these studies are limited to qualitative analysis and hardly provide direct solutions to acoustic design of open ceiling meeting rooms in open-plan offices.
This paper extends the existing research to quantitative analysis of room impulse responses in different frequency bands. An impulse response separation method is proposed, and it is verified with simulations based on the Odeon software and the experiments carried out in 3 different types of rooms. Finally, the proposed method is applied to the Fabpod, a semi enclosed meeting room located in a large indoor office at the Design Research Institute of the RMIT University, to obtain the relative contributions of different acoustic transmission paths to its speech privacy. This method and knowledge obtained can be used by architects to improve the acoustics performance of the next generation Fabpods which are now under design at RMIT University.

The Method
Open ceiling rooms in large offices can be treated as workstations in the acoustic design in open-plan offices. The layout and arrangement of the workstations are important in open-plan office design, while other factors cannot be ignored as well, such as sound absorption, height of screens, degree of workstation enclosure, and room dimensions [5]. The speech signal received in a closed room is the superposition of direct sound and reverberant sound. Reflections arrived within 50 ms after the direct sound are defined as early reflections, which are considered as useful for speech communication while those arrived later are defined as later reflections and are considered as harmful [11]. Thus, the contributions of direct sound and early reflections are considered first.  Typical acoustic paths for sound transmitting from inside to outside a meeting room, where Path 1 is that transmitting through the panel, Path 2 is that diffracting over the panel, Path 3 is that reflecting from the ceiling, and Path 4 is that reflecting from the ceiling and ground.

The theoretical method
The sound pressure level of sound transmitting through the direct path (without panel blocking) depends on the sound power of the sound source and the distance between the source and receiver [11]   2 p, d W 10 where Lw is the sound power level of the sound source and d is the distance between the source and receiver. The sound diffracting over the panel can be obtained with the MacDonald solution [16,17].
where the insertion loss IL can be calculated with where α and ϕ are the angle coordinates of source and receiver in cylindrical coordinates, k is the wave number, R and R' are the distance from the receiver to the source and mirror-image of the source, R1 is the shortest distance from the source to the receiver over the panel, sgn is the signum function and Fr is the Fresnel integral Sound reflecting from the ceiling and ground depends on the absorption coefficient of these surfaces. The sound pressure level after reflecting from a surface is [8]   2 p, refl W 10 r 10 r 1 10 log 4 10 log where αr is the absorption coefficient of the surface and dr is the length of the transmission path. These analytical equations mentioned above will be used in Sections 3 and 4 to verify the reliability of the impulse response separation method.

The impulse response separation method
The impulse responses can be used to describe the transmission properties of a meeting room in an open-plan office, and the properties of the room such as the dimensions of the room, the positions of source and receiver and the existence of physical objects can be estimated with them [18]. Early portion of a room impulse response, which arrives within 50 ms after the direct sound, is beneficial to speech intelligibility while those arrive after 50 ms are harmful. This criterion is often referred to as the early to late energy ratio and is defined by the following equation as [18]     where h(n) is the impulse response and n50 and n51 are the sample numbers corresponding to the 50 th and the 51 st milliseconds after the direct sound.
In general, each peak in the impulse responses corresponds to an acoustic transmission path from the source to the receiver, so the acoustic transmission paths can be separated according to the time delays due to the specific length of the particular path.
Bradley has used this method to mark the components of the impulse response of the initial ceiling reflection path and the initial panel diffraction path [19]. However, only 7 qualitative analysis was carried out in the reference. Here quantitative analysis is proposed in this paper and applied on the room impulse responses. Similar to the early to late energy ratio, the relative sound energy of a particular transmission path to the whole energy ratio can be defined as: where nql and nqu are the sample numbers of the lower and upper limits corresponding to the qth peak component of the impulse response.
Once the physical length of an acoustic transmission path is known, the ratio of sound energy arriving at the receiver within the time interval can be obtained from Eq.
(7). Because the peaks in the measured impulse responses usually are not ideally narrow, a time interval is necessary to obtain the peak energy from the measured impulse responses (4 ms is used in the paper for the system with a sampling frequency of 48 kHz).
So the upper and lower limits of the impulse response in Eq. (7) are defined as: where lq is the length of the qth acoustic transmission path, c0 is the speed of sound in the air, fs is the sampling frequency of the system that is used to measure the impulse responses. In actual calculation, rounding is applied to Eq. (8) as sample numbers are integers. Then the peak component of the impulse responses corresponding to the qth transmission path can be obtained by: where RN(n) is a rectangular windowing function which can be defined as and N is the length of the rectangular windowing function.
Eq. (7) is an expansion of the early to late energy ratio.
where * means convolution, wj is a band pass filter designed for the jth octave band.
This calculation method contains a separation in time domain by using rectangular window functions and another separation in frequency domain by using band pass filters.
Once the total sound pressure level is known, the contributions of each particular transmission path is obtained by where SPL0, j means the sound pressure level of the jth octave band measured at the receiver.
For the impulse response measured in an open-plan office, the relative contributions of each particular acoustic transmission path can be obtained from Eq. (12). Table 1 lists the steps of the proposed impulse response separation method. Table 1 Steps of the impulse response separation method.

Contributions of each acoustic path to STI
STI is a physical quantity representing the transmission quality of speech with respect to intelligibility, and it takes various factors such as reverberation, echoes and interfering noise into account [12]. STI is based on the concept of Modulation Transfer Function (MTF), which was introduced in 1973 [20]. The modulation reduction factor m at modulation frequency F caused by reverberation can be obtained using the impulse response [21,22]: where t is the time, h(t) is the impulse response, and F is the modulation frequency. Eq.
(13) is valid only when the carrier signal is white noise. However, it can be regarded as a good approximation when the carrier signal is an octave band signal and the mrev takes the same value for each carrier signal frequency band [22]. An additional modulation reduction factor is caused by the background noise, which can be obtained by [21]: where SNR is the speech to noise ratio. Thus, the modulation reduction factor is expressed in terms of reverberation and SNR by combining Eqs. (13) and (14)       rev n , , , center frequencies of octave bands fj (125, 250, 500, 1000, 2000, 4000 and 8000) Hz [11].
The apparent signal to noise ratio can be calculated by [11] 10 10log 1 The new SNR without a particular transmission path is obtained by subtracting this portion of the sound pressure from the total sound pressure: With the new impulse responses and the SNRs, the modulation reduction factor 2. Fig. 3 shows main acoustic paths for sound transmitting from inside to outside the meeting room.
11 Figure 2 The schematic diagram of the simulation model.  Table 2 to approximate rigid smooth walls while others remained at 1.0 to further investigate the sound reflecting from the ceiling. Third, the absorption coefficient of the ground was adjusted to the same as the ceiling, so that the 12 reflections from the ground can be included. Finally, the absorption coefficients of all office surfaces were set at those in Table 2 to investigate the reverberation effects of the open-plan office.  As mentioned in Section 2.3, both reverberation and background noise affect STIs.
It is obvious that the removal of acoustic transmission paths changes the property of the room, thus leading to the change of the impulse response and the speech to noise ratio.
The new impulse responses and SNRs without each transmission path are obtained by using Eqs. (18) and (19), and the corresponding STIs are obtained by using Eqs. (13) to (17). The calculated STIs corresponding to different room conditions and different background noise are shown in Table 3.  Fig. 4(a), there is no background noise so the reverberation is the main factor that influences the values of STI and speech privacy. The calculated results are shown in the first column of

Experiments
The In Fig. 6, the impulse responses measured are sufficient clear to be separated and the method proposed in this paper can be used to obtain the contributions of each acoustic path. Considering each transmission path in detail, the sound diffracting over the panel decays with the frequency increasing, which agrees with the trend predicted by the theory.
In contrast, sound reflecting by the ceiling and ground does not agree well with the ideal conditions due to the complex environments in actual offices. The impulse response separation method introduced in this paper provides a convenient way to obtain the contribution of single transmission path in complex actual offices. Besides, it also shows that low frequency sound tends to diffract and dissipate more easily than that of high frequencies, which agrees with the physical principles.
contributions of these paths can be obtained by applying the proposed impulse response separation method. It can be used to calculate the STI according to the method in Section 2.3. The results are shown in Table 4. As discussed in Section 3.1, lack of mask noise makes the effect of different transmission paths on STI weak. In Table 4, the values of STI just change about 0.02 with different transmission paths under consideration. For Office 1, which is a large empty office, the sound diffracting over the panel is the greatest contributor to STI while sound reflecting from the ceiling is the second most important one. For Office 2 and 3, reflecting from the ground is the most important transmission path. Meanwhile, removing the transmission path of sound reflecting from the ceiling even causes the rise of STI in smaller rooms. It is not strange as previous researches have shown that both increasing and decreasing the reflections after direct sound may cause the rise of STI [22]. So removing some transmission paths may be beneficial to the speech intelligibility and the acoustic design should be based on the actual offices to achieve good speech privacy. To sum up, the method introduced in this paper is confirmed to be useful and can provide meaningful knowledge to the acoustic design of offices.

Case study on the Fabpod
The Fabpod shown in Fig. 7(a) is a semi enclosed meeting room located in a large indoor open-plan office, which has non-rectangular overall geometry, non-parallel surfaces and highly articulated interior surface made from an aggregate structure composed of hyperboloid cells with different types of materials [23]. A schematic 18 diagram of the positions of the microphones (R1-R10) and loudspeaker (S1) in the Fabpod is shown in Fig. 7 there is an entrance of it. Sound that arrives earlier than the diffraction over panels is considered as sound transmitting through the panel. As the panels are quite high and close to the ceiling, the zone that sound reflecting from the ceiling can reach is limited, as shown in Fig. 8(b), thus the initial reflection from the ceiling does not exist at some locations.
(a) (b) Figure 8 Acoustic paths, (a) Typical acoustic transmission paths in the Fabpod, where Path 1 is that transmitting through the panel, Path 2 is that diffracting over the panel, Path 3 is that reflecting from the walls, and Path 4 is that reflecting from the ground, (b) Paths of reflecting from the ceiling, where possible path is that can arrive at receivers after reflecting from the ceiling, and impossible path that cannot arrive at receivers.
For the receivers far away from the entrance or walls, such as R3 and R10 in Fig.   7(b), the measured impulse responses are shown in Fig. 9. It is the simplest condition and 20 the main transmission paths are that transmitting through the panels, diffracting over the panel and reflecting from the ground. The peaks in the impulse responses that correspond to these paths are marked in Fig. 9. For the receivers near the wall, such as R1 and R8, sound scattering from the wall become stronger, as shown in Fig. 10. Here the first scattering is additionally marked while subsequent scattering are obvious in later time delay compared to the measured impulse response in Fig. 9. For the receivers near the wall and the entrance, such as R5 and R6, the sound diffracting from the entrance should be considered, which arrives before sound diffracting over the panel. The measured impulse responses are shown in Fig. 11. According to the proposed impulse response separation method, the SPL of each acoustic path of the Fabpod at each receiver are calculated and shown in Table 5    Laying absorption materials on the ceiling and ground can decrease the STI. Sound scattering from the walls is the second important contributor while the receivers are near the wall, and moving the Fabpod to an empty office and away from the wall can improve the STI. Diffracting from the panel is another important path and increasing of the height 23 and width of panel shall decrease its contribution and improves the STI. The entrance is not necessarily a vital concern as it does not have much influence when the receiver is away from the entrance. Besides, sound transmitting through the panel exits but is very weak and can almost be ignored. On the other hand, removing a single transmission path does not affect significantly when the background noise is low. All these factors are important to achieve better speech privacy performance in open-plan offices.

Conclusions
An impulse response separation method is proposed to investigate the contributions