DSpotter Offline Test Tool V2 使用說明
繁體中文 ● 简体中文 ● English
Download
Release notes
注意事項 :
  • 一次辨識複數音檔(.wav)時，需要使用音檔腳本(.spt)，腳本為 UTF-16 little endian 編碼。
  • 腳本每一行必須含有預期辨識音檔的絕對路徑。
  • 每一行的音檔後方為預期辨識的結果，可為單一指令、複數指令，若不填則表示預期無結果。
  • 音檔路徑與後方的預期結果必須以 Tab(非空白字元)分開，若有複數個指令則必須用斜線(/)分開每一個指令。
  • 複數指令的順序同音檔內的指令順序，若腳本填寫"Hello/turn on/turn off"，則預期該音檔的辨識結果也是相同的順序。
開啟模型：
  • 由選單 [檔案] > [載入...] 開啟要測試的模型。
使用者介面：
  • 1. 聲音品質估值: 點選時，於辨識結果後方呈現聲音品質估值的量化數據。
  • 2. 選擇噪音: 點選後，可選擇載入預設噪音或使用者自訂的噪音檔案。
  • 3. 訊號雜訊比: 點選噪音後，可以選擇載入的噪音訊號雜訊比。
  • 4. 載入音檔: 選擇載入預期辨識的音檔(.wav)。
  • 5. 載入腳本: 選擇載入預期辨識的音檔腳本(.spt)。
  • 6. 匯出日誌: 辨識完後可以點選匯出日誌，將辨識結果輸出到文字檔案中。
載入多重群組:
  • DSpotter Offline Test Tool 支援多重群組合併辨識， 由選單 [檔案] > [載入多重群組...]。
  • 所有群組必須是來自同一專案。
  • 首先選擇資料夾。
  • 再來選擇要合併辨識的群組及順序。
  • 選擇的多重群組將會顯示在介面上。
MaxTime
  • DSpotter Offline Test Tool 支援設定 MaxTime，由選單 [選項] > [設定 MaxTime]。
  • MaxTime is 是指令的最長長度, 500(frames) = 5秒。
辨識結果(Result):
  • HitRate: 該指令的正確辨識率。
        該指令被正確辨識的總數 / 該指令預期被辨識的總數。
  • FR(False Reject): 該指令的錯誤拒絕率。
        (該指令預期被辨識的總數 - 該指令被正確辨識的總數) / 該指令預期被辨識的總數。
  • FA(False Alarm): 該指令被錯誤辨識的總數。若使用載入腳本進行辨識且預期無結果，或使用載入音檔進行辨識，辨識的結果屬於 FA。
  • AvgScore: 該指令被正確辨識的平均聲音品質估值。只有在點選聲音品質估值的情況下才會計算 AvgScore。若使用載入腳本進行辨識且預期無結果，或使用載入音檔進行辨識，辨識的分數不計入 AvgScore。
        該指令被正確辨識的總分數(聲音品質估值) / 該指令預期被辨識到的總數。
Download：
Release notes
2025/04/02:
For more compatible with high-dpi displayers, we change our framework from .Net Framework to .Net Core 8.0.
You may need to install .Net Desktop Runtime 8.0 at 1st time executing the HighDPI App. (2.2.21.X)

If your can't install the Runtime, install the .Net Framework version App. (2.2.20.X)
Please keep [DSMT / DSpotter OnlineTestTool / DSpotter OfflineTestTool] in the same version (HighDPI or not).

Date Version(App / Engine / HighDPI App) Description
2025/06/06 2.2.22.0 / 2.2.22.2 / 2.2.22.100 1. Update engine.
2.Add Level information in result.
2025/04/02 2.2.20.10 / 2.2.20.2 / 2.2.21.1 1.Fix: Crash while enable RIR simulation with filename contains non-ANSI characters.
2. Add: Show detailed statistics checkbox.
3. Fix: Fail loading model with very old version.
4. Modify: UI adjustment.
5. Fix: Crashes while some files in .spt does not exist.
2025/02/19 2.2.20.6 / 2.2.20.2 1.Fix: Crash while the wav files too short in Spt sometimes.\n
2025/01/07 2.2.20.5 / 2.2.20.2 1.Fix: Crash while wav file too large.\n
2025/01/03 2.2.20.4 / 2.2.20.2 1.Fix: Can not get result if both Extra Ending and Ending Silence are set to Maximum value.
2024/08/08 2.2.20.3 / 2.2.20.2 1.Update: Engine 2.2.20.2.
2. Fix: Misplacement of result while using LoadSpt in some cases.
3. Modify: Output result shows starting / ending time.
4.Modify: UI adjustment.
2023/11/30 2.2.19.0 / 2.2.19.0 1.Fix: Can not save testing spt if input spt contains no references.
2. Update: Engine 2.2.19.0
2023/10/25 2.2.18.7 / 2.2.18.3 1.Add: Option DotProd Type.
2.More Msgbox for asking saving log.
3.Update: engine 2.2.18.3
2023/10/02 2.2.18.3 / 2.2.18.2 1.Add: AGC SetMaxGain slide-bar.
2.Update: More friendly Msgbox.
3.Update: engine 2.2.18.2
2023/09/01 2.2.18.1 / 2.2.18.0 1.Generate .spt file while "Save TestWav" is checked.
2023/08/22 2.2.18.0 / 2.2.18.0 1.Update engine.
2023/07/04 2.2.17.6 / 2.2.17.0 1.Fix: Fail to check update sometimes.
2023/06/30 2.2.17.5 / 2.2.17.0 1.Fix: Result calculation error.
2023/06/30 2.2.17.4 / 2.2.17.0 1.Fix: Folder of self-add-noise not logged correctly.
2.Fix: Result calculation problem.
3.Fix: SNR calculation problem while adding noise.
4.Fix: Tiny adjust for RIR algorithm.
2023/02/10 2.2.17.3 / 2.2.17.0 1.Fix: Percentage not update correctly.
2023/02/10 2.2.17.1 / 2.2.17.0 1.Fix: Cooldown not working for LoadSpt at the first time.
2023/02/09 2.2.17.0 / 2.2.17.0 1.Update engine
2.Add: Options for Cooldown
3.Fix: Count error for commands end with "^N" sometimes
2022/10/21 2.2.16.2 / 2.2.16.2 1.Fix: RIR simulations crashes on some computers
2.Load model failed if group.mod is not named as "Group_N.mod"
3.Update engine
4.Engine version will be displayed in "Help"->"About"
2022/09/28 2.2.16.1 1.Fix: wrong command energy (=0).
2.Modify: increase running performance while adding noise
2022/09/22 2.2.16.0 1.Add: Room Impulse Response (RIR) simulation.
2.Engine update: 2.2.15.3.
2021/12/03 2.2.14.6 1.Modify: Default FFT algorithm changed to Complex.
2.Fix: FFT algorithm not set properly during the first execution.
2021/11/01 2.2.14.5 1.Fix: Slow recognition speed
2.Fix: UI process percentage may not update immediately
2021/10/27 2.2.14.4 1.Add: Options for FFT algorithm (Real or Complex)
2.Modify: Result format.
2021/03/08 2.2.12.2 1.Fix: Hitrate and FA.
2.Fix: May read error value at ending part of each wav.
2020/12/30 2.2.12.1 1.Fix: Hitrate and FA.
2020/12/29 2.2.12.0 1.Engine Update(AGC for SD)
2.Fix: May not work under Europian language (due to comma and digit).
3.Fix: No FA count while loading .wav.
2020/10/27 2.2.11.0 1.Engine Update(AGC)
2.Fix: non-NTFS system error
3.Fix: big-endian error
4.Fix: 37hr overflow problem
2020/09/23 2.2.10.1 Fix: command number mismatch problem
2020/09/18 2.2.10.0 Engine Update, UI Update
2020/08/20 2.29 Engine Update
2020/07/21 2.28 Engine Update
2020/07/16 2.27 Engine Update
2020/06/15 2.251 Add model memory usage
2020/06/02 2.25 Engine Update
2020/05/25 2.241 Fix potential error
2020/05/25 2.24 Engine Update
2020/05/15 2.231 Add more log
2020/05/13 2.23 Engine Update
2020/05/04 2.221 Fix UI error
2020/04/24 2.22 Engine Update
2020/03/31 2.20 Engine Update
2020/03/05 2.11 Fix add noise error
2020/02/20 2.10 Engine Update
2020/02/07 2.04 Engine update
2020/02/03 2.03 Engine update
2020/01/02 2.022 Fix UI error
2019/12/31 2.021 Fix default value file
2019/12/19 2.02 Display command energy
2019/11/25 2.01 Engine update
2019/11/11 2.00 First release