Projects

It is a two-layer heterogeneous communication network, in which the upper layer can support high-speed transmission through OFDM, and the lower layer can fulfill multi-user access in the network through CDMA. In the system, we can transmit and receive real-time video on two terminals which are connected with broads through network cables using the TCP protocol. The transmitting between the two layers is fulfilled through a gateway or a router.

We design the scheme of the system and implement each part through C language. By Code Composer Studio (CCS), we transplant the system in VS to our hardware platform which is composed of four high-speed signal processing boards. The FPGA chip completes digital up conversion (DUC), digital down conversion (DDC) and filtering. The DSP chip TMS320C6455 from TI can complete processing on baseband and interaction on network layer. The complete procedure is that 

1) the computer at initial terminal collects the video, and passes compressed data packets through the network cable to the lower-layer transmitter, which modulates data into CDMA signals and sends them to receiver;

2) the lower-layer receiver performs real-time demodulation, and then forwards the data packets through the router to the upper-layer transmitter, which modulates data into OFDM wireless signals and sends them to receiver;

3) the upper-layer receiver performs real-time demodulation of signals, and lastly sends the data packets through the network cable to the computer at the other terminal;

4) the computer conducts video decoding and then displays the real-time video which is collected by the computer at initial terminal.

The system can recognize gestures if the object is moving or static and its speed as well as distance. We get gesture signals from changing channel state information (CSI), since this information describes how the signal is propagated from the transmitter through the air to the receiver, reflecting the factor of the signal on each transmission path, such as scattering, fading, power decay with distance, and so on.

On the physical layer, we design the LTE system to transmit and receive signals. The object with various gestures generates changing CSI. At the receiver, we accomplish synchronization and channel estimation using the synchronization and reference signals, and collect data packets. The collected data are composed of noise, objective gesture signals and other reflecting signals such as from the trees or desks.

To extract intended gesture signals without noise and other interference, we smooth the sampled data in a short period and a long period, and then calculate the difference value which contains the phase of the CSI.  The long-period smooth can pick up the signals of static paths in the channel, while the short-period smooth can eliminate the noise.

Further, we find that the slope between phase and frequency is the distance of the object. Hence, we record data at each frequency, fit the preprocessed data linearly, and then compute the difference of adjacent data. In our experiment, since the complex environmental factors such as radio station signals would be mixed in the collected data, some abnormal points appear in the linearly-fitting result. To solve this problem, we utilize the K-means algorithm to cluster data after difference. Then, we drop the abnormal cluster, and calculate the distance of the object with the data in retained cluster. 

*This project is still in progress. We will do more experiment to improve the accuracy and the ability to resist interference.