Reverse it and have Better Communication Quality

Gee‐Pinn Too - z8008070@email.ncku.edu.tw
Yu‐Hao Hsieh
Chih‐Hao Chou
Bo‐Hsien Wu
No. 1 Univ. Rd.,
Dept. of Systems and Naval Mechatronic Eng.,
Natl. Cheng Kung Univ.
Tainan, Taiwan

Sony Lee
Industrial Technol. Res. Inst., Tainan, Taiwan

Popular version of paper 1pSP15
Presented Monday afternoon, May 18, 2009
157th ASA Meeting, Portland, OR

Communication is a universal thing. We always need to get in touch with others, but sometimes we have trouble sending or receiving messages clearly in a noisy environment. To decrease or even to avoid the interference of the noise, a variety of methods and theory have been applied to this research area, such as array signal processing technique. In the present application, array microphones are used and a time-reversal process is introduced to recover the original signal at the source location. This means a clearer signal with high Signal-to-Noise ratio is achieved and audio signal separation is also implemented.

To illustrate the time reversal process, Fig. 1 shows the procedure of the operation. Let us imagine that a signal is coming from a source, transmitting through the medium (inclduing several paths because of the reflective boundary), and received by a receiver (such as a microphone). We can observe there are two delayed versions of signal due to the reflection received after the original one. Then the received signal's time sequence is reversed and transmitted back through the same medium (or the same paths), and received at the source location. The time delay of the delayed version of signal is compensated after back propagation. It’s necessary to reverse the time sequence of the signal received at the source location once more, and finally we acquire a similar signal to compare with the original one. If we have more than one sensor or receiver, we set up a measurement array and administer a time reversal process for each array element. The result will be an inversed version of superposition of signals received at the source location. The more array elements are used, the more similar the level obtained.

Figure 1. Time reversal process diagram

From the above foundation, it’s easy to observe that the most similar result will occur at the source location. But, what if there is more than one sound source? The following simulation will illustrate this case. Figure 2 shows the dispose of three sources and 29 the array elements, and boundary conditions of simulation.

Figure 2. Environmental parameters for simulation

There are three sources in this simulation case. All of sources are cut from different songs. Source no.1 is sung by a lady, while (source no.2 and source no.3) are sung by an identical man. Since there are several different sounds in the simulation space at the same time (include three sources and their reflections), the receivers have no ability to filter but mix all sounds. For example, at microphone no.1, the received signal sounds disordered. After time reversal process, we observe the result at each source location as shown in the following link: source no.1(TR), source no.2(TR), and source no.3(TR). Comparing these results with what microphone no.1 is receiving, we can hear each source song clearly at its original place, respectively. In other words, sound is separated by the procedure.

How can we obtain the signal at the source location without accurately realize the time reversal process? The modeling of path is crucial to the solution. If we could get information about modeling the path that sound propagates, it’s practicable to create a “virtual” path in order to simulate the signal at the source instead of really propagating the reversed signal.

To test the effect of the concept, we set up an experiment in an anechoic room. Fig.3 shows the arrangement of the experiment. The four-inch speaker is used as the source and transmits a section of Chinese pop song sung by a lady (the original source). There is an array made up by 15 microphones, and one reference microphone is set close to the speaker. After obtaining enough information for modeling the path, a level white noise is added to the received signal. In the present study, a Signal-to-Noise ratio is about 20dB, which is a very noisy environment, and hardly any song can be heard (received signal with loud noise). However, the content of the song can be heard fairly well with some existing background noise after the time reversal process (calculated result).

Figure 3. Arrangement of the experiment

    In our opinion, this “virtual” path technique seems useful for real applications because it only requires simple equipment to be realized. Building on these concepts and achievements, the method can be applied on mobile phone or laptop, or even a multimedia room. Someday you’ll hear a more distinct speech when you use your phone! (Work is supported by the joint project of NCKU and ITRI with project No. FY98 A-1)

Figure 4. Diagrams of Application of array microphones and time reversal process on a laptop and a mobile phone