AU2021107071A4

AU2021107071A4 - Radar based non-contact detection of sleep stage

Info

Publication number: AU2021107071A4
Application number: AU2021107071A
Authority: AU
Inventors: Changyang Li; Zhiyu Ning
Original assignee: Rudder Technology Pty Ltd
Current assignee: Rudder Technology Pty Ltd
Priority date: 2021-08-24
Filing date: 2021-08-24
Publication date: 2021-12-02
Anticipated expiration: 2029-08-24
Also published as: AU2021107071B4

Abstract

SA non-contact detection system of sleep state, in operable communication with a radar sensor. The radar sensor continuously collects I/Q channel signals and combines the IQ channel signals as single-channel signals. The single-channel signal is used to measure the heartbeat rate and breath rate using the frequency distribution method. An autoencoder is then invented to extract the intrinsic features of the single-channel signal, which maintain the important information but filter the unexpected noise. The extracted features, heartbeat rate and breath rate together form the predictor for the detection of the sleep stage. An estimator, comprising several convolutional layers and a softmax function, is invented to detect the specific sleep stage from awake, REM (rapid eye movement), light sleep and deep sleep. Drawings 106 Estimator Data y 108 105 Autoencoder 103 109 110 104 Radar Sensor Memory Detector10 1 101 Figure 1 104 RadarSensor 112 1 channel data Q channel data 113 Single Channel 114 Signal 117 115/\/ Heartbeat Rate Autoencoder Breath Rate 116 Intrinsic 118 Features Predictor 119 Estimator 106 Figure 2

Description

Drawings

108 Data y 106 Estimator

103 109 105 Autoencoder 110 104 Radar Sensor Memory

Detector10 1

101 Figure 1

104 RadarSensor

112 1channel data Q channel data 113

Single Channel 114 Signal

117

115/\/ Heartbeat Rate Autoencoder Breath Rate 116

Intrinsic 118 Features

Predictor 119

Estimator 106

Figure 2

Radar based non-contact detection of sleep stage

Field of the Invention

[0001] This invention relates to non-contact detection of sleep stages, in operable communication with radar sensors.

Background of the Invention

[0002] Smart health monitoring is more and more popular in aged caring communities. Sleep during the night is an important factor that reflects people's health conditions. The sleep status can be generally classified into four stages, including awake, REM (rapid eye movement), light sleep and deep sleep. In most cases, these four stages usually occur during a full sleep cycle and a complete sleep contains several sleep cycles. Therefore, identifying the sleep stages and calculating the duration of each stage in a complete sleep are important in smart health monitoring systems.

[0003] In the current market, wearing a smartwatch is becoming a benchmark to detect the specific sleep stage. Smartwatches could measure a person's heartbeat rate and stability, and then combine these measurements to infer which stage the person is in during sleep. Although the precision of smartwatches is quite high for sleep stage detection, this contact way impedes its development among those people who do not like or want to wear matches during their sleep. Therefore, a non-contact way to detect sleep stages is desirable.

[0004] This invention will present a new approach that uses radar sensors to achieve non contact detection for the sleep stage.

Summary of the Disclosure

[0005] The sleep stage detection system comprises a radar sensor, an autoencoder and an estimator.

[0006] The radar sensor collects I and Q channel signals and then combines the two-channel signal into the one-channel signal.

[0007] The heartbeat rate and breath rate are estimated from the one-channel signal.

[0008] The autoencoder is designed to extract intrinsic features that maintain the most important information but filter unexpected noise.

[0009] The extracted features, together with heartbeat rate and breath rate, form the predictor of the sleep stage.

[0010] The estimator, comprising several convolutional layers and a softmax function, is invented to detect the specific sleep stage (awake, REM, light sleep or deep sleep).

Brief Description of the Drawings

[0011] Figure 1 shows an exemplary radar-based sleep stage detection system in accordance with an embodiment.

[0012] Figure 2 shows exemplary data processing in the system of Figure 1 in accordance with an embodiment.

Description of Embodiments

[0013] Figure 1 shows a heartbeat rate detector 101 for a home environment, bedroom room or the like.

[0014] Detector 101 comprises several sensors 104, autoencoder 105 and estimator 106. The data 103 is operable in system memory 102 for interpretation and execution of the computational functionality.

[0015] In the embodiment shown, the radar devices are installed above ceiling 107.

[0016] The detector can be configured to detect the breath rate of a person 111 who is laying on bed 110 within a supervision area 109.

[0017] In a specific time t, the radar sensor 104 would collect two kinds of channel signal, i.e., I channel signal 112 (denoted as I(t)) and Q channel signal 113 (denoted as Q (t)). The two-channel signal can be further processed as single-channel signal 114 using the following equation:

[0018] [Equation 1] S(t) = arctan Q M)

[0019] During a time period T(ti, t 2 ,..., ta ),a series signal can be obtained:

[0020] [Equation 2] S(T) = [S(t), S(t 2 ), , S(tn)]

[0021] Similar to the existing methods, the heartbeat rate 115 and breath rate 116 in the time period T can be measured by the frequency distribution method, denoted as HR (T) and BR(T), respectively.

[0022] Then, an autoencoder 117 is invented to extract the intrinsic features 118 of S(T). The detailed structure of the autoencoder 117 is shown in the table below.

Layer Input Output Index dimension dimension

encoding 1 A 4096 layers 2 4096 2048

3 2048 1024

decoding 4 1024 2048 layers 5 2048 4096

6 4096 A

[0023] Denote the output of the third layer, with S(T) being input, as S(T). S(T) is another kind of representation of S(T), which maintains the intrinsic information of S(T) but filters unexpected noise.

[0024] Then, S(T), HR(T) and BR(T) can be together formed as the predictor 119, denoted as PD(T), for the sleep stage during the time period T:

[0025] [Equation 3] PD(T) = [S(T), HR(T), BR(T) ]T

[0026] An estimator 106 is invented to perform the classification of different sleep stages including awake, REM, light sleep and deep sleep. The estimator is composed of several convolutional layers and a softmax function. The detailed structure of the estimator is shown in the table below.

LayerIndex Filter Size Filter Stride Pad Filter Channel

1 3x3 1 1 512

2 3x3 1 1 512

3 3x3 1 1 512

4 3x3 1 1 512

5 3x3 1 1 1024

6 (fully - 1024 connected)

7 (fully - 4 connected)

8 (softmax function)

[0027] The loss function below is used to train the estimator:

[0028] [Equation 4]: L = Ik * log(pk)

[0029] where k = 0,1,2,3 represents the different sleep stages (awake, REM, light sleep and deep sleep) respectively. The lk = 1 if the person is in lk stage, otherwise 0. The Pk is the k the element of the output of the estimator. The training aim is to iteratively update the filters of the estimator to minimize the loss function of Equation 4.

[0030] After a well-trained estimator 106 is obtained, the sleep stage g(T) can be detected using the following equation:

[0031] [Equation 5]: g(T) = argmax(pk)

[0032] The foregoing description, for purposes of explanation, used specific nomenclature to provide a thorough understanding of the invention. However, it will be apparent to one skilled in the art that specific details are not required in order to practice the invention. Thus, the foregoing descriptions of specific embodiments of the invention are presented for purposes of illustration and description. They are not intended to be exhaustive or to limit the invention to the precise forms disclosed as obviously many modifications and variations are possible in view of the above teachings. The embodiments were chosen and described in order to best explain the principles of the invention and its practical applications, thereby enabling others skilled in the art to best utilize the invention and various embodiments with various modifications as are suited to the particular use contemplated. It is intended that the following claims and their equivalents define the scope of the invention.

[0033] The term "approximately" or similar as used herein should be construed as being within 10% of the value stated unless otherwise indicated.

Claims

Claims

) 1. Non-contact detection of sleep stage, in operable communication with a radar sensor.
2. The system as claimed in claim 1, wherein the detector comprises a radar sensor, an autoencoder and an estimator.
3. The system as claimed in claim 1 and 2, wherein the radar sensor collects I/Q channel signals for the detection of heartbeat rate and breath rate.
4. The system as claimed in claim 1 and 2, wherein the autoencoder is used to extract the intrinsic features of radar signals that maintain the important information but filters the unexpected noise.
5. The system as claimed in claim 1 and 2, wherein the estimator is used to detect the specific sleep stage from heartbeat rate, breath rate and the extracted features.