201021767 六、發明說明： 【發明所屬之技術領域】 本發明是有關於-種腦波分析方法，特別是指一種利 用單通道腦波信號分析睡眠狀態的睡眠生理信號分析方法 0 【先前技術】 現代人發生失眠、睡眠週期障礙，及睡眠呼吸中止症 等睡眠疾病的比例越來越高。臨床上對於睡眠疾病的診斷 ’醫生主要針對是否有「良好的睡眠品質」與「完整的睡 眠結構」評估。其中，「良好的睡眠品質」屬主觀感受，通 常透過問卷方式進行評估，而「完整的睡眠結構」屬客觀 事實’則須仰賴儀器領測生理信號後判讀。目前醫學檢測 機構最常使用多重睡眠檢查儀（polysomnography，簡稱 PSG )針對受測者蒐集整晚睡眠期間的腦電圖（多通道）、 眼動圖（雙通道）、心電圖、肌電圖、胸腹部活動、口鼻氣 流、血壓、血液含氧量等信號。 然而，PSG設備龐大、操作複雜、成本高昂，僅有大 型醫療院所或健檢中心提供檢測服務，病患接受檢測需在 院内過夜。然而，在院内過夜的睡眠情境與日常不同，加 上眾夕感測器干擾導致難以自然入眠，降低檢測效果。此 外，PSG設備所測得一整晚的數據量十分龐大，需有經驗 豐富的醫師大量閱讀後仔細推敲判讀。 為克服上述問題，近幾年來已有針對檢測裝置或數據 分析方法提出各種改良方案。例如中華民國發明第i288875 201021767 號專利「多重信號分類方法」，主要是針對電腦運算量大及 運算時間長的問題進行改善，其利用小波轉換求得信號能 量並藉此進行分類，所得到的結果可協助醫生等專業人士 在解讀仏號時縮短時間。但由於此技術所採生理信號仍需 來自PSG等大型設備，因此僅適用於醫院。 此外，還有中華民國發明第129926〇號專利「多重生理 監測墊」，提供一具有感測器的軟墊以偵測受測者的生理信 號’藉此達到受測者身體零感測的目的。此技術雖可能應 用於居家檢測，但其獲得的信號與現有臨床數據完全不同❹ 該等數據無法供醫生等專業人士進行判讀及診斷。 由以上可知，在所獲得的信號數據可供醫學臨床使用 的條件下，現有改良方案僅能儘可能針對腦波等生理信號 正理數據以協助醫生判讀，對於患者而言，受檢測時因環 境及儀器干擾難以入眠的問題仍無法解決。因此，最根本 的解決方案應是發展一套可用於居家檢測的睡眠生理信號 檢測裝置，以便受測者能自行在家進行檢測，獲得真實的 數據，且該檢測裝置所獲得之數據及電腦分析結果需可供❹ 醫生等專業人士參考。 【發明内容】 因此，本發明之目的，即在提供一種僅使用單通道腦 波信號即可分析得到準確的睡眠分期結果且適用於居家檢 測的睡眠生理信號檢測暨分析裝置。 本發明之另一目的，在於提供一種僅使用單通道腦波 信號即可分析得到準確結果的睡眠生理信號分析方法與系 201021767 統。 於疋 本發明睡眠生理信號檢測暨分析裝置包 系2,及一睡眠生理信號分析系統。該信號梅取系5 用模組、一傳輸模組及一储存模組。感測模組 〜1單通道腦波㈣’或更同時彳貞測眼動信號及心電 仏號:傳輸模組將感測模組所獲得的數位信號傳至睡眠生 理信號刀n統進行分析。料模㈣謂數位信號儲存201021767 VI. Description of the invention: [Technical field to which the invention pertains] The present invention relates to a method for analyzing brain waves, in particular to a method for analyzing sleep physiological signals using a single-channel brain wave signal for analyzing sleep state. [Prior Art] Modern The proportion of sleep disorders such as insomnia, sleep cycle disorders, and sleep apnea is increasing. Clinical diagnosis of sleep disorders 'Doctors mainly evaluate whether there is a "good sleep quality" and a "complete sleep structure." Among them, "good sleep quality" is subjective and is usually assessed by questionnaire. The "complete sleep structure" is an objective fact. It depends on the instrument to take the physiological signal and then interpret it. At present, medical testing institutions often use polysomnography (PSG) to collect EEG (multi-channel), eye movement (two-channel), electrocardiogram, electromyography, and chest during sleep. Abdominal activity, nose and mouth airflow, blood pressure, blood oxygen levels and other signals. However, PSG equipment is large, complicated to operate, and costly. Only large medical institutions or health check-up centers provide testing services, and patients need to be tested overnight in the hospital. However, the sleep situation in the hospital overnight is different from the daily routine, and it is difficult to naturally fall asleep and reduce the detection effect by adding interference to the sensor. In addition, the amount of data measured by the PSG equipment for a whole night is very large, and it requires a large number of experienced physicians to read and carefully interpret the interpretation. In order to overcome the above problems, various improvements have been proposed for detection devices or data analysis methods in recent years. For example, the "Multiple Signal Classification Method" of the Chinese Patent No. i288875 201021767 is mainly for improving the problem of large computer computation and long operation time, and uses wavelet transform to obtain signal energy and classify it by using the result. It can help doctors and other professionals to shorten the time when interpreting nicknames. However, because the physiological signals collected by this technology still need to come from large equipment such as PSG, it is only suitable for hospitals. In addition, there is also the Chinese Patent No. 129926, "Multiple Physiological Monitoring Pad", which provides a cushion with a sensor to detect the physiological signal of the subject, thereby achieving the purpose of zero measurement of the subject's body. . Although this technique may be applied to home testing, the signals obtained are completely different from the existing clinical data. These data cannot be interpreted and diagnosed by professionals such as doctors. It can be seen from the above that under the condition that the obtained signal data can be used for medical clinical use, the existing improved scheme can only assist the doctor in reading the physiological signal rational data such as brain waves, and for the patient, the environment and The problem that the instrument interferes with difficulty in sleeping can still not be solved. Therefore, the most fundamental solution should be to develop a set of sleep physiological signal detection devices that can be used for home detection, so that the testee can perform self-testing at home, obtain real data, and the data obtained by the test device and computer analysis results. Need to be used by professionals such as doctors. SUMMARY OF THE INVENTION Accordingly, it is an object of the present invention to provide a sleep physiological signal detecting and analyzing apparatus which can analyze an accurate sleep staging result using only a single channel brain wave signal and is suitable for home detection. Another object of the present invention is to provide a sleep physiological signal analysis method and system 201021767 which can analyze an accurate result using only a single channel brain wave signal. Yu Yu The sleep physiological signal detection and analysis device package 2 of the present invention, and a sleep physiological signal analysis system. The signal is a module, a transmission module and a storage module. Sensing module ~ 1 single-channel brain wave (four) 'or at the same time speculating eye movement signal and ECG nickname: the transmission module transmits the digital signal obtained by the sensing module to the sleep physiological signal knife for analysis . Material module (4) is called digital signal storage
本發月睡眠生理仏號}析系统針對接收到的單通道腦 波錢利用慢波、α波’及紡錘波（spindles)進行睡眠分 期刀析包括一前置處理模組、一 α波偵測模組、一慢波 偵測模組、—紡錘波偵測模組、一眼動信號分析模組及 一分期計算模組。 前置處理模組將接收到的腦波原始訊號，首先需去除 雜讯，接著進行濾波。原始信號經五組不同頻帶濾波器濾 波後’可得到慢波（〇 5〜4Ηζ )、Θ波（2〜7Ηζ )、^波（ 7〜12Hz)、纺錘波（u~16Hz)與高頻波（3〇~4〇Ηζ)五種 頻帶信號。 在慢波的彳貞測及分析方面’慢波分析模組先從慢波頻 帶化號中找出符合R&K睡眠分期規則（目前醫界採用的臨 床睡眠分期規則）所定義（0.5〜4Hz且振幅大於75//V)之 波开> ’接著计算出每時段的腦波中慢波出現時段所佔比例 ’並重建整夜的慢波百分比以供後續分期演算。前述時段 ，例如以30秒為—時段。 201021767 在α波偵測及分析方面由於0波、“波與纺錘波的 頻帶接近α波刀析模組是同時比較前述三種波的頻帶作 號每秒的均方根振幅，找出以α波為主要成分的秒數，^ 積30秒後’可找出心波為主要成分的秒數佔該3〇秒中 的比例（α% )，接著可畫出整夜的α波百分比。 在紡錘波的偵測及分析方面，紡錘波偵測模 組是利用This month's sleep physiology } } 析 针对 针对 针对 针对 针对 } } } } } } } } } } } } } } } } 针对 针对 针对 针对 针对 } } } } 针对 } } } } } } } } } } } } } The module, a slow wave detection module, a spindle wave detection module, an eye signal analysis module and a stage calculation module. The pre-processing module will receive the original signal of the brain wave, first need to remove the noise, and then filter. The original signal is filtered by five different sets of band filters to obtain slow waves (〇5~4Ηζ), chopping waves (2~7Ηζ), ^ waves (7~12Hz), spindle waves (u~16Hz) and high-frequency waves ( 3〇~4〇Ηζ) Five frequency band signals. In the analysis and analysis of slow wave, the slow wave analysis module first finds the R&K sleep staging rule (currently the clinical sleep staging rules adopted by the medical community) from the slow wave band number (0.5~4Hz). And the amplitude is greater than 75//V) and then 'calculates the proportion of the slow wave occurrence period in the brainwaves per period' and reconstructs the percentage of slow waves throughout the night for subsequent stage calculations. The aforementioned time period is, for example, 30 seconds as a time period. 201021767 In the alpha wave detection and analysis, due to the zero-wave, "wave and spindle wave frequency band close to the alpha wave knife-analysis module is to simultaneously compare the rms amplitude of the three wave frequency bands per second, find out the alpha The number of seconds after the wave is the main component, after 30 seconds, the number of seconds in which the heart wave is the main component can be found in the ratio of the 3 seconds (α%), and then the percentage of the alpha wave can be drawn overnight. Spindle wave detection module is used for the detection and analysis of spindle waves.
Schimicek演算法偵測得紡錘波。由於紡錘波具有特殊的波 形與長度，Schimicek演算法是將峰對峰振幅大於閥值（以 25 μν為例）的片段定義為第一階段紡錘波。由於R&K規則❹ 疋義紡錘波長度僅0.5〜2秒，Schimicek演算法將符合波長 疋義之波形找出，定義為第二階段紡錘波。為避免所偵測 到的紡錘波是其他頻帶之雜訊，Schimicek演算法計算第二 階段之紡錘波前後各2.5秒之a頻帶與高頻頻帶（3〇〜4〇 Hz) 均方根振幅。若α波與紡錘波均方根振幅的比值大於1 2， 則將此紡鐘波視為雜訊而予以刪除。高頻之均方根振幅若 大於5 μν ’亦將紡錘波視為雜訊而予以刪除，雜訊處理後 剩餘的紡錘波定義為最後階段紡錘波。每30秒為一段的腦 © 波出現紡錘波的個數定義為紡錘波密度（spindles density)， 並建立整夜的紡錘波密度變化。 依據先前計算出每段的慢波比例、α波比例，及纺錘波 密度，先利用瀘、波器進行平滑化處理後得到三種波整夜變 化趨勢圊形，接著.由分期計算模組進行睡眠分期演算法。 由於醫學臨床上睡眠分期，共分為清醒期、非快速眼 動（non-rapid eye movement，簡稱 NREM)第一期、第二 201021767 期、第三期、第四期與快速眼動（REM)期等六期。清醒 期的信號包括兩類，一類在信號上無明顯特徵，完全為受 測者活動之雜訊；另一類則是入眠前閉上眼睛的時候，腦 波幾乎為（超過50%)規律的α波。因此，本發明之睡眠分期 演算法首先找出波比例大於5〇%的時期，當作清醒期； β波比例小於50%者，作以下之進一步區分。. 由於非快速眼動第三期腦波出現低頻（約0.5〜4ΙΙζ)且大 振幅（>75 μν)的慢波或稱§波（delta wave)佔20〜5〇〇4;非快 速眼動第四期慢波比例更多（大於50%)，該第三與第四期可 合稱慢波睡眠（slow wave sleep)或深眠（deep sleep)，因此找 出慢波比例大於20%者作為深眠期。 由於紡錘波會出現於非快速眼動第二、三、四期，其 他時期紡錘波密度幾乎為零；因此，經由前述判斷剩餘的 信號中’有出現紡錘波者應為非快速眼動第二期，又可稱 淺眠（light sleep)期。至於未出現紡錘波者，應為快速眼動 期或非快速眼動第一期，只要再輔以利用眼動信號感測器 擷取到的眼動信號’由眼動信號分析模組利用波形變化的 斜率、振幅及寬度等參數且設定閥值的方式偵測快速眼動 時期，即可以作進一步區分；或者，由於非快速眼動第一 期為睡眠暫態轉變現象’所佔時間非常短暫，較簡單的做 法也可忽略，例如把快速眼動期及非快速眼動第一期整體 視為快速眼動期，並將分類在淺眠期之前的快速眼動期直 接修正為清醒期。 此外’本發明的信號擷取系統的感測模組更可包括一 201021767 心電信號感測器，且睡眠生理信號分析系統更可包括一心 電信號分析模組，藉由偵測心電信號中的R波並計算r波 與R波間的時間間期數列，且將此數列帶入希爾柏轉換% Hilbert transform)式中，最後利用閥值的設定即可算出 呼吸中止次數，判斷是否有睡眠呼吸障礙。 本發明睡眠生理信號分析方法，僅需以單通道腦波信 號做為輸人資料，所計算獲得的睡眠分期結果與傳統利用 腦波、眼動圖、下顎肌電圖、心、電信號等多重信號所分析 的睡眠分期結果十分接近’且若除了單通道腦波信號之外 ，輔以參考單通道眼動信號及心電信號，即可得到完整睡 眠分析報表。整體觀之，使用本發明睡眠生理信號分析方 法，感測器對睡眠受測者的干擾可降到最低，且由於所需 的生理信號減少，本發明睡眠生理信號檢測暨分析裝置十 分適合作為居家檢測用；又’本發明分析得到的睡眠分期 結果十分簡單易懂’可降低人工解讀信號的時間及避免誤 差；長期獲得的睡眠分期結果更可提供給醫生或專業人士 進行判斷且作為進一步的診療的依據。 【實施方式】 有關本發明之前述及其他技術内容、特點與功效，在 以下配合參考圖式之一個較佳實施例的詳細說明中，將可 清楚的呈現。 參閲圖1，本發明睡眠生理信號檢測暨分析裝置1〇〇之 較佳實施例包含一信號擷取系統2,及一睡眠生理信號分析 系統3。仏號摘取系統2包括一感測棋組21、一偟私播 201021767 22及一儲存模組23。The Schimicek algorithm detects the spindle wave. Since the spindle wave has a special shape and length, the Schimicek algorithm defines a segment with a peak-to-peak amplitude greater than the threshold (in the case of 25 μν) as the first-stage spindle wave. Since the R&K rule 疋 纺 纺 spindle length is only 0.5~2 seconds, the Schimicek algorithm will find the waveform corresponding to the wavelength , meaning, defined as the second stage spindle wave. In order to avoid the detected spindle wave being noise in other frequency bands, the Schimicek algorithm calculates the rms amplitude of the 2.5-band a-band and high-frequency band (3〇~4〇 Hz) before and after the second stage of the spindle wave. If the ratio of the alpha wave to the root mean square amplitude of the spindle wave is greater than 12, the spinning clock is removed as a noise. If the rms amplitude of the high frequency is greater than 5 μν ', the spindle wave is removed as noise, and the remaining spindle wave after the noise processing is defined as the final stage spindle wave. The brain is a section of the brain every 30 seconds. The number of spindle waves is defined as the spindle density, and the spindle wave density change is established overnight. According to the previous calculation of the slow wave ratio, the α wave ratio, and the spindle wave density of each segment, the smoothing process of the 泸 and the waver is used to obtain three kinds of wave all-night change trend shape, and then by the stage calculation module. Sleep staging algorithm. Due to the clinical clinical sleep stage, it is divided into awake period, non-rapid eye movement (NREM) first period, second 201021767 period, third period, fourth period and rapid eye movement (REM). Period six. There are two types of signals during the awake period. One type has no obvious characteristics on the signal, and it is completely the noise of the subject's activity. The other type is when the eyes are closed before going to sleep. The brain waves are almost (more than 50%) regular α. wave. Therefore, the sleep staging algorithm of the present invention first finds a period in which the wave ratio is greater than 5〇%, and considers it as a waking period; if the β wave ratio is less than 50%, the following further distinction is made. Because of the low-frequency (about 0.5~4ΙΙζ) and large amplitude (>75 μν) slow wave or delta wave accounted for 20~5〇〇4; non-rapid eye movement The fourth phase of slow wave is more (more than 50%), and the third and fourth phases can be called slow wave sleep or deep sleep, so find the slow wave ratio is greater than 20%. As a deep sleep period. Since the spindle wave will appear in the second, third and fourth phases of non-rapid eye movement, the spindle wave density is almost zero in other periods; therefore, the remaining signal in the above judgment should be the non-rapid eye movement second. Period, can also be called the light sleep period. As for the case where the spindle wave does not appear, it should be the first phase of rapid eye movement or non-rapid eye movement, as long as it is supplemented by the eye movement signal captured by the eye movement signal sensor. Change the slope, amplitude and width parameters and set the threshold to detect the rapid eye movement period, which can be further distinguished; or, because the non-rapid eye movement first period is the sleep transient transition phenomenon, the time taken is very short The simpler method can also be ignored. For example, the rapid eye movement period and the non-rapid eye movement first period are regarded as the rapid eye movement period, and the rapid eye movement period classified before the light sleep period is directly corrected to the waking period. In addition, the sensing module of the signal acquisition system of the present invention may further comprise a 201021767 ECG signal sensor, and the sleep physiological signal analysis system may further comprise an ECG signal analysis module for detecting the ECG signal. R wave and calculate the time interval between the r wave and the R wave, and bring this sequence into the Hilbert transform formula, and finally use the threshold setting to calculate the number of breath suspensions to determine whether there is sleep. Respiratory disorder. The sleep physiological signal analysis method of the invention only needs to use the single-channel brain wave signal as the input data, and the calculated sleep staging result and the traditional use of brain wave, eye movement picture, sacral electromyogram, heart and electric signal, etc. The sleep staging results analyzed by the signal are very close to 'and, in addition to the single-channel brainwave signal, supplemented by the reference single-channel eye-motion signal and ECG signal, a complete sleep analysis report can be obtained. Overall, using the sleep physiological signal analysis method of the present invention, the interference of the sensor to the sleep subject can be minimized, and the sleep physiological signal detecting and analyzing device of the present invention is very suitable as a home due to the required physiological signal reduction. For detection; and 'the sleep staging result obtained by the invention is very simple and easy to understand' can reduce the time of manual interpretation of signals and avoid errors; the long-term obtained sleep staging results can be provided to doctors or professionals for judgment and as further diagnosis and treatment. Basis. The above and other technical contents, features, and advantages of the present invention will be apparent from the following detailed description of the preferred embodiments. Referring to Fig. 1, a preferred embodiment of the sleep physiological signal detecting and analyzing apparatus of the present invention comprises a signal capturing system 2 and a sleep physiological signal analyzing system 3. The nickname extracting system 2 includes a sensing chess set 21, a private broadcast 201021767 22 and a storage module 23.
本實施例之感測模組21包括一組單通道腦波感測器（ 圖未示）、一組單通道眼動信號感測器（圖未示），及一組 '^電k號感/則器（圖未不）。以腦波感測器舉例來說，可採 用單通道貼片式感測器，其類比電路規格頻率響應為 〇·5〜50 Hz、輸入阻抗為h6〜10 οΩ、共模拒斥比為ι〇〇犯 與放大倍率ιο,οοο倍，數位電路以100 Hz取樣頻率與8 bit 解析度將信號數位化，但不以前述規格為限。貼片黏貼位 置是：正極位於前額右眼上方、負極位於左耳後乳突處， 參考電極則由傳統頭顱位置改至軀幹肋骨中線處，以利量 測其他生理信號，例如當作心電信號感測器。單通道眼動 信號感測器可偵測水平眼動信號，貼片黏貼位置是兩眼眼 尾水平向外約一公分位置。 傳輸模組22使用例如藍芽技術，即時將感測模組21 所獲得的數位信號傳至睡眠生理信號分析系統3進行分析 。儲存模組23用以將數位信號儲存。 本發明睡眠生理信號分析系統3包括_前置處理㈣ 31、1波福測模組32、一慢波债測模組33、一坊鐘波债 測模組34、-眼動信號分析模組％、一心電信號分析模植 37 ’及一分期計算模組35。 配合參閱圖2，在步驟s + ® «te ^ 你7邵、中，前置處理模組31將接 到的腦波原始訊號，首先愛土 需去除雜訊，接著以零相位減浊 器濾波。原始信號經五組不 。 丨J頻帶慮波盗；慮波後，可棋丨 0.5〜4Hz之慢波、2〜7Hz之0你7 , ^皮、7〜12Hz之α波、丨丨〜^❿ 201021767 之纺鐘波與30~40Hz之尚頻波五種頻帶信號。接著，分三 頭進行步驟S21〜s41、步驟S22〜s42，及步驟s23〜s43。 在步驟Sn中，慢波分析模組33從慢波頻帶信號中找 出符合R&K睡眠分期規則所定義之波形一〇·5〜4Hz且振幅 大於75 β V ^本實施例是利用越零分析偵測慢波：以數位信 號取樣點數值正負號轉變後的第一個取樣點作為越零點， 相鄰越零點之間的半波頻率與半波振幅符合前述慢波定義 者為慢波。 接著，步驟Sn以【式一】計算每3〇秒為一段的腦波© 中慢波所佔比例，以供後續分期演算。 1 n SW〇/〇 = 30 ^ durationk ) x 1 〇〇% 【式一】 其中，η是慢波出現的次數。【式一】將每一次（k)出 現慢波的持續時間（durati〇n)相加，再除以一時段腦波的 時間’得到該段腦波中慢波所佔比例（下稱慢波百分比， SW% )。The sensing module 21 of the embodiment includes a set of single-channel brain wave sensors (not shown), a set of single-channel eye-motion signal sensors (not shown), and a set of '^ electric k senses / then (Figure is not). For example, a brain wave sensor can be a single-channel chip sensor with an analog circuit frequency response of 〇·5~50 Hz, an input impedance of h6~10 οΩ, and a common mode rejection ratio of ι. 〇〇 与 and magnification ιο, οοο times, the digital circuit digitizes the signal at a sampling frequency of 100 Hz and 8 bit resolution, but is not limited to the aforementioned specifications. The placement position of the patch is: the positive electrode is located above the right eye of the forehead, the negative electrode is located at the posterior mastoid of the left ear, and the reference electrode is changed from the traditional skull position to the midline of the trunk rib to measure other physiological signals, for example, as a heart. Electrical signal sensor. The single-channel eye movement signal sensor can detect the horizontal eye movement signal, and the position of the patch is about one centimeter outward from the tail of the two eyes. The transmission module 22 transmits the digital signal obtained by the sensing module 21 to the sleep physiological signal analysis system 3 for analysis using, for example, Bluetooth technology. The storage module 23 is configured to store digital signals. The sleep physiological signal analysis system 3 of the present invention comprises: _ pre-processing (four) 31, 1 wave test module 32, a slow wave debt test module 33, a square clock wave debt test module 34, an eye movement signal analysis module %, an ECG signal analysis model 37' and a staging calculation module 35. Referring to Figure 2, in step s + ® «te ^ you 7 Shao, middle, pre-processing module 31 will receive the original brainwave signal, first love the soil to remove the noise, then filter with zero phase opacity . The original signal passes through five groups.丨J band wave pirates; after the wave, you can play a slow wave of 0.5~4Hz, 0~7Hz 0 you ^, ^ skin, 7~12Hz alpha wave, 丨丨~^❿ 201021767 spinning clock wave and Five frequency band signals of 30~40Hz frequency. Next, steps S21 to s41, steps S22 to s42, and steps s23 to s43 are performed in three steps. In step Sn, the slow wave analysis module 33 finds the waveform defined by the R&K sleep staging rule from the slow wave band signal, and the amplitude is greater than 75 β V ^. The embodiment uses zero Analyze and detect the slow wave: the first sampling point after the sign of the digital signal sample point is converted to the zero point, and the half-wave frequency and the half-wave amplitude between the adjacent zero-crossing points are the slow wave defined by the slow wave. Next, the step Sn calculates the proportion of the slow wave in the brain wave © for every 3 sec., for the subsequent stage calculation. 1 n SW〇/〇 = 30 ^ durationk ) x 1 〇〇% [Formula 1] where η is the number of occurrences of slow waves. [Equation 1] Add the duration of the slow wave (durati〇n) every time (k), and divide by the time of the brain wave in a period of time to get the proportion of the slow wave in the brain wave (hereinafter referred to as the slow wave). Percentage, SW%).
Q 最後，步驟s41重建整夜的慢波百分比得到如圖3所示 之圖形。 在皮伯測及分析方面，由於0波、α波與纺鐘波的 ^帶 波刀析模組32是同時比較前述三種波的頻帶 h號每秒的，方根振幅而從中找出心皮。本實施例α波债 m、.且32疋利用【式二】分別計算三種波的頻帶信號每秒 :其中RMS代表均方根（細mean sq霞）值 ，N代移、令的信號點數，Xi代表每個信號點的振幅 10 201021767 (amplitude) °Q Finally, step s41 reconstructs the percentage of slow waves throughout the night to obtain a graph as shown in FIG. In the aspect measurement and analysis, the 0-wave, alpha-wave and the spinning-wave wave-wave-splitting module 32 simultaneously compares the frequency of the three kinds of waves, h-number per second, and finds the carpel from the square root amplitude. . In the present embodiment, the alpha wave bonds m, . and 32 疋 use [formula 2] to calculate the frequency band signals of the three waves per second: where RMS represents the root mean square (fine mean sq) value, and the number of signal points of the N generation shift and the order , Xi represents the amplitude of each signal point 10 201021767 (amplitude) °
RMS amplitudeRMS amplitude
【式二】 若第k秒〇：波的均方根振幅較0波及紡錘波者大代 表該秒以α波為主要成分（步驟 S22)，即【式三】中Sk= 1 在步驟S32中’利用【式三】累積3〇秒後，可找出^ 參 波為主要成分的秒數佔該30秒中的比例（α % )。 1 30 4=跑)祕·..··.··.【式三】 接著，由步驟S42畫出整夜的α波百分比如圖4所示。 在纺鐘波的偵測及分析方面，紡錘波偵測模組34在步 驟S23是利用Schimicek演算法偵測得紡錘波。由於紡錘波 具有特殊的波形與長度’ Schimicek演算法是將峰對峰振幅 大於閥值（以25 μν為例）的片段定義為第一階段紡錘波。由 於R&K規則定義紡錘波長度僅〇.5〜2秒，Schimicek演算法 將符合波長定義之波形找出，即第二階段紡錘波^為避免 所偵測到的紡錘波是其他頻帶之雜訊，Schimicek演算法計 算第二階段之紡錘波前後各2.5秒之α頻帶與高頻頻帶（3〇〜 40 Hz)均方根振幅。若α波與紡錘波均方根振幅的比值大於 1.2，則將此紡錘波視為雜訊刪除。高頻波之均方根振幅若 大於5 μν，亦將紡錘波視為雜訊濾除，雜訊處理後的紡錘 波為最後被確定的紡鐘波。 11 201021767 每 丰又的腦波出現紡錘波的個數定義為紡錘波 密度（SP_eS density)(步綠I3)，整夜的纺錘波密度變化 如圖5所示（步驟s43)。 依據先前計算出每段的慢波比例、α波比例，及紡錘 波密度’進行平滑化後得到三種波整夜變化趨勢圖形接 著由分期計算模組35進㈣眠分期演算法（步驟S5〜步驟 S8)。 在步驟s5中，本實施例之睡眠分期演算法首先找出^ 波比例大於50%的時期，當作「清醒期」。 ❹ α波比例小於50%者，在步驟心作進一步區分：由於 非快速眼動第三期腦波出現低頻（約〇5〜4Hz)且大振幅（>75 μν)的慢波佔20〜50%’非快速眼動第四期慢波大於5〇0/。’ 本實施例採用將第三與第四期合併當作深眠（deep sleep)期 的理論（美國睡眠醫學會於2007年建議合併稱之為慢波睡 眠）’找出慢波比例大於20%者作為「深眠期」。 找出深眠期之後，剩餘的由步驟S7繼續分析：由於紡 錘波會出現於非快速眼動第二、三、四期，其他時期紡錘© 波密度幾乎為零；因此，經由前述判斷剩餘的信號中，有 出現紡錘波者應為非快速眼動第二期，又稱「淺眠（light sleep)期」。 至於未出現紡錘波者’應為快速眼動期或非快速眼動 第一期，本實施例在步驟S8中，是利用感測模組21中眼動 信號感測器所測得的眼動信號，由眼動信號分析模組36以 波形變化的斜率、振幅及寬度等參數且設定閥值的方式偵 12 201021767 測快速眼動時期’即可區分出快速眼動期與非快速眼動第 一期。詳言之，以圖6波形來說，本實施例偵測快速眼動 的方式’是以滿足以下條件者視為快速眼動期的波形：丨波 峰前的斜率大於波峰後的斜率；2.波谷和波峰距離小於〇.5 秒；3.兩波峰間距大於〇 5秒；4振幅及斜率的閥值。 特別說明的是，步驟S5〜Ss的順序不以本實施例為限， 也可以如圖7所示，先偵測紡鐘波開始出現且連續達兩分 鐘的開始點，該開始點之前的時期，進一步觀察α波百分 比大於50%的時間點，由於α波要在眼睛閉上後才會顯現 出來，因此第一次大於50%阿法波出現之前，判斷為眼睛 還沒閉上的「清醒期」’因此第〜欠以>5()%之前歸類為 凊醒期」’其餘為「非快速動眼第一期」；紡錘波開始的 時間點之後，觀察慢波比例是否大於，若是則屬「深 眠期」’右否則以前述快速眼動偵測方法找出「快速眼動期 」，並以紡錘波出現與否找出非快速動眼第二期也就是「 淺眠期」；若皆否’則為「非快速眼動第一期」。 利用前述分期步驟，最後可繪得如圖8所示的睡眠分 期圖，以供醫生或專業人士參考。 此外，爲獲得完整的睡眠分析結果，本實施例還利用 心電信號分析模組37找出心電信號中的R波並計算r波 與R波間的時間間期數列’且將此數列帶人希爾柏轉換式 中’利用閥值的設；t找出呼吸中止時間點，最後算出整夜 呼吸中止次數。 综上所述，本發明利用單通道腦波信號，輔以單通道 13 201021767 眼動信號’建立-種簡易、簡便且可靠的睡眠生理信號分 析方法及H適合居家使用以獲得長時間、符合實際睡 眠狀態的㈣’且分析結果呈現方式符合醫學臨床使用的 規格，有助於醫生或專業人士·，確實能達成本發明之 目的。 惟以上所述者，僅為本發明之較佳實施例❿已，當不 能以此限定本發明實施之範圍，即大凡依本發明申請專利 範圍及發明說明内容所作之簡單的等效變化與修飾，皆仍 屬本發明專利涵蓋之範圍内。 ◎ 【圖式簡單說明】 圖1疋一系統方塊圖，說明本發明睡眠生理信號檢測 暨分析裝置的較佳實施例； 圖2是本發明睡眠生理信號分析方法之較佳實施例的 流程圖； 圖3是利用該實施例偵測慢波後繪出的整夜慢波百分 比的變化圖； 圖4是該實施例偵測α波後繪出的整夜〇波百分比的 © 變化圖； 圖5是該實施例偵測紡錘波後繪出的整夜紡錘波密度 的變化圖 圖6是該實施例中偵測快速眼動的示意圖； 圖7是本發明另一睡眠分期的流程圖；及 圖8是利用該實施例得到的睡眠分期圖。 14 201021767 【主要元件符號說明】 100 ·····. ••睡眠生理信號檢 33. 慢波偵測模組 測暨分析裝置 34. 紡錘波偵測模組 2 ......... •信號擷取系統 35. 分期計算模組 21........ •感測模組 36· 眼動信號分析模組 22........ •傳輸模組 37· 心電信號分析模組 〇 Ί........ C •儲抒模組 Si 應波 3 ......... •睡眠生理信號分 S21 〜S23 · 偵測波形 析系統 S31 〜s33. 計算百分比或密度 31·..·..·· •前置處理模組 S41 〜S43. 繪出整夜變化圖形 32........ • α波偵測模組 s5〜 -Sg ··· 分期演算 15[Equation 2] If the kth second 〇: the root mean square amplitude of the wave is larger than the 0 wave and the spindle wave, the α wave is the main component for the second (step S22), that is, [S3] in the formula 3 is in step S32. 'Using [Formula 3] for 3 seconds, you can find out the ratio of the number of seconds in which the reference wave is the main component to the 30 seconds (α % ). 1 30 4=Run) Secret·..·····. [Formula 3] Next, the percentage of alpha waves that are drawn overnight by step S42 is as shown in FIG. In the detection and analysis of the spinning clock wave, the spindle wave detecting module 34 detects the spindle wave using the Schimicek algorithm in step S23. Since the spindle wave has a special waveform and length, the Schimicek algorithm defines a segment whose peak-to-peak amplitude is greater than the threshold (in the case of 25 μν) as the first-stage spindle wave. Since the R&K rule defines a spindle wave length of only 55~2 seconds, the Schimicek algorithm will find the waveform according to the wavelength definition, that is, the second stage spindle wave is to avoid the detected spindle wave being miscellaneous in other frequency bands. The Schimicek algorithm calculates the rms amplitude of the 2.5-band alpha band and the high-frequency band (3 〇 to 40 Hz) before and after the second stage of the spindle wave. If the ratio of the alpha wave to the root mean square amplitude of the spindle wave is greater than 1.2, the spindle wave is considered as noise removal. If the rms amplitude of the high-frequency wave is greater than 5 μν, the spindle wave is also considered as noise filtering, and the spindle wave after the noise processing is the last determined spinning clock wave. 11 201021767 The number of spindle waves per brain wave is defined as the spindle wave density (SP_eS density) (step green I3), and the overnight spindle wave density changes as shown in Fig. 5 (step s43). According to the previous calculation, the slow wave ratio, the α wave ratio, and the spindle wave density of each segment are smoothed to obtain three kinds of wave overnight trend patterns, and then the staging calculation module 35 enters (four) sleep stage algorithm (step S5 to step) S8). In step s5, the sleep staging algorithm of the present embodiment first finds a period in which the ratio of the wave is greater than 50%, and regards it as the "awake period". ❹ The ratio of α wave is less than 50%, and the step is further differentiated: due to non-rapid eye movement, the third wave of brain waves appears low frequency (about 〜5~4Hz) and the large amplitude (>75 μν) slow wave accounts for 20~ 50% 'non-rapid eye movement fourth phase slow wave is greater than 5〇0/. 'This example uses the theory that the third and fourth phases are combined as a deep sleep period (American Sleep Medicine Association proposed to merge in 2007 called slow wave sleep)' to find that the slow wave ratio is greater than 20%. As a "deep sleep period." After finding out the deep sleep period, the remaining analysis is continued from step S7: since the spindle wave will appear in the second, third and fourth phases of non-rapid eye movement, the spindle © wave density is almost zero in other periods; therefore, the remaining Among the signals, the occurrence of spindle waves should be the second phase of non-rapid eye movement, also known as the "light sleep period". As for the case where the spindle wave does not appear, it should be the rapid eye movement period or the non-rapid eye movement first stage. In the embodiment, in step S8, the eye movement measured by the eye movement signal sensor in the sensing module 21 is used. The signal is determined by the eye movement signal analysis module 36 by the parameters such as the slope, the amplitude and the width of the waveform change and the threshold value is set. 12 201021767 The rapid eye movement period can be distinguished to distinguish the rapid eye movement period from the non-rapid eye movement level. One issue. In detail, in the waveform of FIG. 6, the method for detecting rapid eye movement in this embodiment is a waveform that is regarded as a rapid eye movement period when the following conditions are satisfied: the slope before the chopping peak is larger than the slope after the peak; The valley and crest distance is less than 〇5 seconds; 3. The two peak spacing is greater than 〇5 seconds; 4 the amplitude and slope threshold. In particular, the order of steps S5 to Ss is not limited to this embodiment, and as shown in FIG. 7, the start point of the spinning clock wave starting to appear for two minutes, the period before the starting point, may be detected first. Further observe the time point when the alpha wave percentage is greater than 50%. Since the alpha wave is to be revealed after the eye is closed, the first time before the occurrence of more than 50% of the alpha wave, the eye is still not closed. "Therefore, the first ~ owed to > 5 ()% before classified as the awake period" and the rest as "non-rapid eyes first period"; after the start of the spindle wave, observe whether the slow wave ratio is greater than, if It is a "deep sleep period". Right, the "rapid eye movement period" is found by the above-mentioned rapid eye movement detection method, and the second phase of non-rapid eye movement is called "light sleep period" by the presence or absence of a spindle wave; If it is no, it is "the first phase of non-rapid eye movement". Using the aforementioned staging steps, a sleep staging map as shown in Figure 8 can be drawn for reference by a doctor or professional. In addition, in order to obtain a complete sleep analysis result, the embodiment also uses the ECG signal analysis module 37 to find the R wave in the ECG signal and calculate the time interval sequence between the r wave and the R wave' and bring the series to the person. In the Hiller conversion type, 'use the threshold value; t find the time point of the breathing stop, and finally calculate the number of times the whole night stops. In summary, the present invention utilizes a single-channel brain wave signal, supplemented by a single channel 13 201021767 eye movement signal 'establishment - a simple, simple and reliable sleep physiological signal analysis method and H suitable for home use to obtain long time, in line with reality The sleep state of (four) 'and the analysis results presented in accordance with the specifications of medical clinical use, to help doctors or professionals, can indeed achieve the purpose of the present invention. However, the above is only the preferred embodiment of the present invention, and the scope of the present invention is not limited thereto, that is, the simple equivalent change and modification according to the scope of the present invention and the description of the invention. All remain within the scope of the invention patent. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing a preferred embodiment of a sleep physiological signal detecting and analyzing device of the present invention; FIG. 2 is a flow chart showing a preferred embodiment of the sleep physiological signal analyzing method of the present invention; FIG. 3 is a graph showing the change of the percentage of the entire night slow wave after the slow wave is detected by using the embodiment; FIG. 4 is a graph showing the change of the percentage of the overnight chopping wave after the alpha wave is detected in the embodiment; FIG. FIG. 6 is a schematic diagram of detecting rapid eye movement in the embodiment; FIG. 7 is a flow chart of another sleep staging according to the present invention; 8 is a sleep staging chart obtained by using this embodiment. 14 201021767 [Description of main component symbols] 100 ······•• Sleep physiological signal detection 33. Slow wave detection module measurement and analysis device 34. Spindle wave detection module 2 ........ • Signal acquisition system 35. Stage calculation module 21........ • Sensing module 36· Eye movement analysis module 22........ • Transmission module 37· Heart Electric signal analysis module 〇Ί........ C • Storage module Si should wave 3 ......... • Sleep physiological signal points S21 ~ S23 · Detection waveform analysis system S31 ~ s33. Calculate the percentage or density 31·..·..·· • Pre-processing module S41 ~ S43. Draw the whole night change pattern 32........ • α wave detection module s5~ -Sg · ·· Staging calculus 15