“Ataxia” was initially used to describe various uncoordinated characteristics of different diseases, such as gait, movement, heartbeat, etc. Now it is more specifically used to express the symptoms of motor mismatching synchronization and balance disorder after the brain, cerebellum, deep sensation (proprioception), vestibular and other systems are damaged (Bastian, 1997). Different pathological locations often show different characteristics. Sensory ataxia is caused by the impairment of somatosensory nerve, which leads to the interruption of sensory feedback signals and therefore, the body incoordination is caused. For Cerebellar Ataxia patients, the Romberg’s sign was positive, the typical symptoms include walking slowly, rolling, etc. Symptoms were mild when eyes were open and aggravated when eyes were closed (Fadic et al., 1997; Donnelly, 2011). Cerebellar ataxia patients are more common, it is a loss of body muscle coordination caused by cerebellar disease. Trunk ataxia often indicates cerebellar vermis lesions, and limb ataxia often indicates cerebellar hemisphere lesions. The corresponding patients often have eye tremor, low muscle tension, unclear speech, and other symptoms (Diener and Dichgans, 1992; Bastian et al., 1996).

In clinical testing, SA syndrome is very easy to be misdiagnosed as CA syndrome, which leads to the inability of patients with ataxia to get correct diagnosis and treatment in time. At present, several international medical organizations have formed to study ataxia (Klockgether and Paulson, 2011), and several ataxia assessment scales were developed, such as “International Cooperative Ataxia Rating Scale for pharmacological assessment of the cerebella syndrome (ICARS)” (Trouillas et al., 1997), “Scale for the assessment and rating of ataxia (SARA)” (Schmitz-Hübsch, 2006). Some scholars have also done relevant research on the clinical detection and differentiation of SA and CA symptoms, and given the clinical diagnosis method (Chhetri et al., 2014). Both the assessment scale and related research work have referred to two basic indicators: Romberg’s sign and gait; which could be used for Clinical detection and differentiation of SA and CA.

In typical indoor environment, the wireless signal emitted by the transmitter would be affected by the objects or the human body; and the refraction, reflection and diffraction may cause multipath effect. These homologous wireless signals in different propagation paths show different physical characteristics at the receiving end, such as the amplitude and phase of the receiving signals, which contain rich information from the external environment.

When the receiver detects that the signal changes, it indicates that the external environment has been changed. By de-noising the acquired data and further processing with classification algorithm, we can reduce the environmental factors that lead to the change of the received signal, so as to obtain the desirable information in the environment. In this work, since the antennas were used for detection and monitoring applications in a regular shape room, basic omnidirectional monopole antennas were considered; for irregular shape space and room, specially designed antennas would be necessary to enhance the performance and accuracy of sensing. The main differences between received signal strength and channel state information are explained in Zhu and Zhang (2010) and Zheng et al. (2013). CSI considers the number of antennas and subcarriers, and can measure more fine-grained information, the facility which confirms to the IEEE 802.11n standard was used to collect the CSI data. The IEEE 802.11n standard uses orthogonal frequency division multiplexing (OFDM) to transmit a single data stream with 20 MHz bandwidth through 56 orthogonal subcarriers, the signals transmitted on each subcarrier have different signal strength and phase (Lorincz and Begusic, 2006). The facility used in this paper provides 30 available subcarriers to users. Next, we will further explain the principle of C-Band wireless sensing measurement from the formula.

It is known that the channel impulse response (CIR) is generally used to describe the multipath effect in wireless channels. Under linear time-invariant conditions, the CIR can be expressed as follows,

In the formulas above, a_i, θ_i and τ_i represent the attenuation factor, phase shift, and time delay of the i-th path, respectively, N is the total number of propagation paths, and δ(τ) is Dirichlet pulse function.

Since the multipath propagation of signals can cause delay and attenuation, we can also describe the channel by channel frequency response (CFR), as shown in (2),

Where Y is the vector representation of receiving signal, X is the vector representation of transmitting signal, N is the noise matrix, H is the channel attenuation matrix and describes the attenuation factor of signal on each transmission path, the dimension of H can be expressed as:

Where R_N and T_N are the number of receiving antennas and transmitting antennas, respectively.Sub_N is the number of subcarriers.

CSI is essentially a representation of the frequency response of each subcarrier channel, as shown in (4),

In (4),|h(f_i, t)|, arg(h(f_i, t)), andf_i denote the amplitude, phase, and central frequency of i-th subcarrier, respectively.

Since the patient takes some time to perform Romberg’s test and gait detection, we need continuous monitoring, and the received CSI data can be expressed as:

Where D represents the data stream received by the receiving antenna, P_i (i = 1, 2…, n) represents packet. Each packet contains 30 subcarriers, and n is the total number of received packets. D constitutes the analysis data source for detecting and distinguishing sensory ataxia and cerebellar ataxia. Since the phase of subcarriers in each packet is random, this paper will mainly use the amplitude information of subcarriers (Yang et al., 2017).

The experiment was carried out in an approximate ward room, its size is 7 m×5 m.

The experimental equipment includes two industrial control computers equipped with facilities conforming to the IEEE 802.11n standard. The transmitter is equipped with an omnidirectional antenna, and the receiver is equipped with three omnidirectional antennas. Since each antenna receives packets containing 30 subcarriers, we will get 3× 30 subcarriers for each packet at the receiving end, which greatly increases the data size. The experimental scene is shown in Figure 5.

In the experiment, 10 subjects are considered; they are divided into two groups with five people in each group. We set the contract awarding frequency to 200 Hz. For Romberg’s test, we collected a total of 12 s of data, including 6 s for the open eyes stage and 6 s for the close eyes stage. For the gait detection experiment, considering the site constraints and the walking speed between different objects, we collected a total of 5 s of data, the amount of data is enough to distinguish the abnormal gait.

For each subject, Romberg’s test and gait detection were repeated 24 times. We collected three sets of data each day and collected the complete data in about 1 week. There are 120 sets of experimental data for each of the test items for sensory ataxia and cerebellar ataxia. At the same time, we also collected 120 sets of Romberg’s test and gait detection data under normal conditions as a reference.

Due to the noise in the environment, to ensure the credibility and accuracy of the results, the data is processed considering the following steps (Figure 6):

Sensory Ataxia and Cerebellar Ataxia are neurological diseases which affect the patients’ quality of life seriously; therefore, their detection at early stage are very important and necessary. In this paper, non-contact wireless sensing technology has been proposed to discriminate symptoms between the two diseases. The advantages include improvement of comfort, overcoming self-consciousness enhancing, etc. The main merit of the system lies in its convenience and price cost advantage. We firstly preprocess the data by removing outliers, wavelet transform filtering, then data features are extracted, finally, we use BP Neural Network, SVM, RF machine learning algorithms to train the model. The experimental results show that most of the algorithms can achieve more than 96% prediction accuracy, which can effectively discriminate between sensory ataxia and cerebellar ataxia, and prove that the technical scheme described in this paper is effective. Next, we will further explore the application of C-Band wireless sensing technology in healthcare, and propose more clinical application programs to make clinical detection more accurate, reliable and smarter, so as to reduce the burden on clinicians and patients.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the Northwest Women’s and Children’s Hospital, Xi’an Jiaotong University Health Science Center. The patients/participants provided their written informed consent to participate in this study.

QZ: manuscript writing. XZ, YL, and XY: guidance. XY: editing, project management, and funding. QA: verification. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.