## Box Plot: Acoustic Scene Analysis ### Overview The image presents three box plots comparing the performance of different audio processing techniques (BMVDR, BLCMV, BMVDR-N, BLCMV-N) against a Reference and an Anchor. The plots analyze the "Interfering source", "Background noise", and "Complete acoustic scene" based on a "Score" metric. Each box plot displays the distribution of scores for each technique, including the median, quartiles, and outliers. ### Components/Axes * **X-axis:** Categorical, representing the audio processing techniques: "Reference", "Anchor", "BMVDR", "BLCMV", "BMVDR-N", "BLCMV-N". * **Y-axis:** Numerical, labeled "Score", ranging from 0 to 100 in increments of 25. * **Box Plots:** Each box plot represents the distribution of scores for a given technique. The box indicates the interquartile range (IQR), the line within the box represents the median, and the whiskers extend to the furthest data point within 1.5 times the IQR. Outliers are marked as red plus signs (+). * **Titles:** Each plot has a title indicating the acoustic condition being analyzed: "Interfering source", "Background noise", and "Complete acoustic scene". ### Detailed Analysis #### Interfering Source * **Reference:** Score consistently at 100. * **Anchor:** Score consistently at 0. * **BMVDR:** The box spans from approximately 10 to 25, with a median around 17.5. There are outliers near 0 and 35. * **BLCMV:** The box spans from approximately 60 to 75, with a median around 70. There are outliers near 27 and 100. * **BMVDR-N:** The box spans from approximately 25 to 40, with a median around 35. There are outliers near 50. * **BLCMV-N:** The box spans from approximately 65 to 75, with a median around 72. There are outliers near 45 and 92. #### Background Noise * **Reference:** Score consistently at 100. * **Anchor:** Score consistently at 0. * **BMVDR:** The box spans from approximately 15 to 35, with a median around 25. * **BLCMV:** The box spans from approximately 25 to 45, with a median around 37.5. * **BMVDR-N:** The box spans from approximately 65 to 80, with a median around 72.5. * **BLCMV-N:** The box spans from approximately 65 to 80, with a median around 75. #### Complete Acoustic Scene * **Reference:** Score consistently at 100. * **Anchor:** Score consistently at 0. * **BMVDR:** The box spans from approximately 10 to 25, with a median around 17.5. There are outliers near 0 and 10. * **BLCMV:** The box spans from approximately 40 to 75, with a median around 57.5. There is an outlier near 100. * **BMVDR-N:** The box spans from approximately 25 to 50, with a median around 37.5. * **BLCMV-N:** The box spans from approximately 70 to 90, with a median around 75. There is an outlier near 40. ### Key Observations * The "Reference" consistently scores 100 across all three acoustic conditions, while the "Anchor" consistently scores 0. * BLCMV and BLCMV-N generally outperform BMVDR and BMVDR-N in all three conditions. * The performance of BMVDR-N and BLCMV-N is significantly better than BMVDR and BLCMV in the "Background noise" condition. * The "Interfering source" condition shows the most variability in performance across the different techniques. ### Interpretation The box plots provide a comparative analysis of different audio processing techniques under varying acoustic conditions. The "Reference" and "Anchor" serve as benchmarks, representing ideal and baseline performance, respectively. The data suggests that BLCMV and BLCMV-N are generally more effective at handling both interfering sources and background noise compared to BMVDR and BMVDR-N. The addition of noise reduction (indicated by the "-N" suffix) appears to significantly improve performance in the presence of background noise. The variability in scores for the "Interfering source" condition suggests that the effectiveness of these techniques may be highly dependent on the specific characteristics of the interfering source. Overall, the plots highlight the strengths and weaknesses of each technique, providing valuable insights for selecting the most appropriate audio processing method for a given acoustic environment.

\n ## Box Plot Charts: Performance Scores Under Different Acoustic Conditions ### Overview The image presents three box plot charts, each comparing the performance scores of different algorithms ("Reference", "Anchor", "BMVDR", "BLCMV", "BLCMV-N") under three distinct acoustic conditions: "Interfering source", "Background noise", and "Complete acoustic scene". Each chart displays the distribution of scores for each algorithm using box plots and individual data points. ### Components/Axes Each chart shares the following components: * **X-axis:** Algorithm names: "Reference", "Anchor", "BMVDR", "BLCMV", "BLCMV-N". * **Y-axis:** "Score", ranging from 0 to 100, with tick marks at 0, 25, 50, 75, and 100. * **Data Representation:** Box plots representing the interquartile range (IQR) and median of the scores for each algorithm. Individual red data points are plotted on top of the box plots, representing individual observations. Error bars extend from the box plots, indicating the range of the data. * **Titles:** Each chart has a title indicating the acoustic condition: "Interfering source", "Background noise", and "Complete acoustic scene". ### Detailed Analysis or Content Details **Chart 1: Interfering Source** * **Reference:** The median score is approximately 5, with the box extending from roughly 0 to 15. Individual data points are scattered between 0 and 90. * **Anchor:** The median score is approximately 70, with the box extending from roughly 50 to 85. Individual data points are scattered between 20 and 95. * **BMVDR:** The median score is approximately 25, with the box extending from roughly 15 to 40. Individual data points are scattered between 10 and 60. * **BLCMV:** The median score is approximately 65, with the box extending from roughly 50 to 80. Individual data points are scattered between 30 and 90. * **BLCMV-N:** The median score is approximately 75, with the box extending from roughly 60 to 90. Individual data points are scattered between 40 and 95. **Chart 2: Background Noise** * **Reference:** The median score is approximately 5, with the box extending from roughly 0 to 15. Individual data points are scattered between 0 and 60. * **Anchor:** The median score is approximately 75, with the box extending from roughly 60 to 90. Individual data points are scattered between 30 and 95. * **BMVDR:** The median score is approximately 25, with the box extending from roughly 15 to 40. Individual data points are scattered between 10 and 60. * **BLCMV:** The median score is approximately 70, with the box extending from roughly 55 to 85. Individual data points are scattered between 30 and 95. * **BLCMV-N:** The median score is approximately 75, with the box extending from roughly 65 to 90. Individual data points are scattered between 40 and 95. **Chart 3: Complete Acoustic Scene** * **Reference:** The median score is approximately 5, with the box extending from roughly 0 to 15. Individual data points are scattered between 0 and 80. * **Anchor:** The median score is approximately 75, with the box extending from roughly 60 to 90. Individual data points are scattered between 30 and 95. * **BMVDR:** The median score is approximately 25, with the box extending from roughly 15 to 40. Individual data points are scattered between 10 and 60. * **BLCMV:** The median score is approximately 55, with the box extending from roughly 40 to 70. Individual data points are scattered between 20 and 90. * **BLCMV-N:** The median score is approximately 75, with the box extending from roughly 60 to 90. Individual data points are scattered between 40 and 95. ### Key Observations * The "Reference" algorithm consistently performs the worst across all three acoustic conditions, with low median scores and a wide range of data points. * The "Anchor" algorithm generally performs well, with high median scores and relatively tight distributions. * "BLCMV-N" consistently shows high performance, often comparable to or slightly better than "Anchor". * "BMVDR" shows moderate performance, consistently better than "Reference" but generally lower than "Anchor" and "BLCMV-N". * The spread of data points (as indicated by the individual points and error bars) varies significantly between algorithms and acoustic conditions, suggesting varying levels of robustness. ### Interpretation The data suggests that the "Anchor" and "BLCMV-N" algorithms are the most effective at maintaining performance across different acoustic environments. The "Reference" algorithm is consistently the least effective. The "BLCMV-N" algorithm appears to offer a slight improvement over the "Anchor" algorithm, particularly in the "Interfering source" and "Background noise" conditions. The wide spread of data points for some algorithms indicates that their performance is sensitive to specific instances within each acoustic condition. This could be due to variations in the characteristics of the interfering source, background noise, or the overall acoustic scene. The consistent poor performance of the "Reference" algorithm suggests it is not well-suited for these types of acoustic environments. The improvement offered by "BLCMV-N" over "BLCMV" suggests that the "N" component provides a beneficial enhancement, potentially related to noise reduction or interference cancellation. The charts provide a comparative analysis of algorithm performance, allowing for informed selection based on the expected acoustic conditions.

## Box Plots: Method Performance Across Acoustic Scenarios ### Overview The image contains three side-by-side box plots comparing the performance of different audio processing methods across three acoustic scenarios: "Interfering source," "Background noise," and "Complete acoustic scene." Each plot evaluates methods labeled as "Reference," "Anchor," "BMVDR," "BLCMV," "BMVDR-N," and "BLCMV-N." Scores range from 0 to 100, with red dots representing medians, blue boxes showing interquartile ranges (IQR), and red stars indicating outliers. ### Components/Axes - **X-axis (Methods)**: - Reference - Anchor - BMVDR - BLCMV - BMVDR-N - BLCMV-N - **Y-axis (Score)**: 0 to 100 (linear scale). - **Plot Titles**: 1. "Interfering source" 2. "Background noise" 3. "Complete acoustic scene" - **Legend Elements**: - Red dots: Medians - Blue boxes: Interquartile ranges (IQR) - Red stars: Outliers ### Detailed Analysis #### Interfering Source - **Reference**: Highest score (~100), no outliers. - **Anchor**: Lowest score (~0), with 3 outliers (~5, ~10, ~15). - **BMVDR**: Median ~20, IQR ~10–30, 1 outlier (~5). - **BLCMV**: Median ~70, IQR ~60–80, 1 outlier (~90). - **BMVDR-N**: Median ~50, IQR ~40–60, 1 outlier (~70). - **BLCMV-N**: Median ~75, IQR ~65–85, 1 outlier (~95). #### Background Noise - **Reference**: Highest score (~100), no outliers. - **Anchor**: Lowest score (~0), with 2 outliers (~5, ~10). - **BMVDR**: Median ~20, IQR ~10–30, 1 outlier (~5). - **BLCMV**: Median ~30, IQR ~20–40, 1 outlier (~50). - **BMVDR-N**: Median ~25, IQR ~15–35, 1 outlier (~40). - **BLCMV-N**: Median ~70, IQR ~60–80, 1 outlier (~85). #### Complete Acoustic Scene - **Reference**: Highest score (~100), no outliers. - **Anchor**: Lowest score (~0), with 3 outliers (~5, ~10, ~15). - **BMVDR**: Median ~20, IQR ~10–30, 1 outlier (~5). - **BLCMV**: Median ~50, IQR ~40–60, 1 outlier (~70). - **BMVDR-N**: Median ~30, IQR ~20–40, 1 outlier (~50). - **BLCMV-N**: Median ~75, IQR ~65–85, 1 outlier (~90). ### Key Observations 1. **Reference** consistently achieves the highest scores across all scenarios, with no outliers. 2. **Anchor** performs poorly in all scenarios, with frequent outliers. 3. **BMVDR** and **BLCMV** show moderate performance, with BLCMV outperforming BMVDR in "Interfering source" and "Complete acoustic scene." 4. **BMVDR-N** and **BLCMV-N** demonstrate improved scores compared to their non-N variants, but still fall short of Reference. 5. Outliers are most frequent in "Anchor" and "BLCMV-N" (e.g., ~95 in "Interfering source"). ### Interpretation The data suggests that the **Reference** method is the most robust, achieving near-perfect scores without variability. **Anchor** is the least effective, with significant performance gaps and outliers. **BLCMV-N** and **BMVDR-N** show incremental improvements over their base methods, but their performance remains suboptimal compared to Reference. The presence of outliers in "BLCMV-N" and "Anchor" indicates potential instability or edge-case failures. These results highlight the importance of method selection based on acoustic context, with Reference being the gold standard.