Bird Call Identification Using Machine Learning and Acoustic Feature Analysis in MATLAB

Author : Waqas Javaid

Abstract

Bird vocalizations provide valuable information for species identification, biodiversity assessment, and ecological monitoring. This study presents a machine learning-based Bird Call Identifier developed in MATLAB for the automatic recognition of bird species using acoustic signal analysis. Synthetic bird call signals representing different bird species are generated and processed under realistic environmental conditions [1]. A comprehensive feature extraction framework is employed to obtain important acoustic characteristics, including spectral centroid, spectral spread, entropy, spectral flux, roll-off frequency, zero-crossing rate, energy, dominant frequency, spectral flatness, and RMS value [2]. These features are used to construct a discriminative feature vector for classification purposes. A K-Nearest Neighbor (KNN) classifier is trained and tested using normalized feature data to identify bird species accurately [3]. The system performance is evaluated through waveform analysis, spectrogram visualization, frequency spectrum analysis, feature clustering, and confusion matrix assessment. Experimental results demonstrate high classification accuracy and effective separation among bird species in the feature space [4]. The proposed approach offers a computationally efficient, reliable, and scalable solution for automated bird call recognition and bioacoustic monitoring applications. The developed framework can support wildlife conservation, ecological research, and intelligent environmental monitoring systems.

1. Introduction

Birds are important indicators of environmental health and biodiversity, making their monitoring essential for ecological research and conservation efforts.

Figure 1: Bird call identification system showing the detected bird call waveform, signal envelope, predicted species, and overall classification accuracy.

Figure 1 represents the bird species identification methods rely on visual observation and expert knowledge, which can be time-consuming, labor-intensive, and subject to human error. In recent years, advances in digital signal processing and machine learning have enabled the development of automated systems capable of recognizing bird species through their vocalizations [5]. Bird calls contain distinctive acoustic patterns that can be analyzed to identify species accurately, even in environments where visual observation is difficult. Automated bird call recognition has gained significant attention due to its potential applications in wildlife monitoring, habitat assessment, biodiversity conservation, and ecological studies [6]. Acoustic monitoring systems offer a non-invasive and cost-effective approach for collecting information about bird populations over large geographical regions [7]. The increasing availability of computational tools and machine learning algorithms has further enhanced the ability to process and classify complex audio signals efficiently. Feature extraction techniques play a critical role in transforming raw audio recordings into meaningful representations that can be utilized for classification tasks. Spectral and temporal characteristics such as spectral centroid, spectral spread, entropy, energy, dominant frequency, and zero-crossing rate provide valuable information about bird vocalizations [8]. Machine learning classifiers can learn patterns from these features and accurately distinguish between different bird species. Among various classification methods, the K-Nearest Neighbor (KNN) algorithm is widely used due to its simplicity, effectiveness, and robustness in pattern recognition problems. This study presents a MATLAB-based Bird Call Identifier that combines acoustic feature extraction with machine learning classification for automated bird species recognition [9]. Synthetic bird call signals representing multiple bird species are generated and analyzed under realistic conditions. The extracted features are normalized and used to train a KNN classifier capable of identifying unknown bird calls. Various visualization techniques, including waveform analysis, spectrograms, frequency spectra, feature clustering, and confusion matrices, are employed to evaluate system performance. The proposed framework demonstrates how signal processing and machine learning can be integrated to develop an intelligent bioacoustic monitoring system [10]. The developed approach is computationally efficient, scalable, and suitable for both academic research and practical environmental monitoring applications. The results highlight the effectiveness of acoustic feature analysis in distinguishing bird species and provide a foundation for future advancements in automated wildlife monitoring technologies.

1.1 Background of Bird Monitoring

Birds are important components of natural ecosystems and serve as indicators of environmental health. Their presence, abundance, and behavior provide valuable information about biodiversity and habitat quality [11]. Monitoring bird populations helps researchers understand ecological changes and conservation needs. Traditional monitoring methods often require extensive fieldwork and expert knowledge. Therefore, automated approaches are increasingly being explored for efficient bird identification.

1.2 Importance of Bird Vocalizations

Birds communicate through various vocalizations that contain species-specific acoustic patterns. These vocal signals are often easier to capture than visual observations, especially in dense forests and remote environments. Bird calls provide information about species identity, mating behavior, territorial activities, and migration patterns [12]. Acoustic monitoring has emerged as a powerful tool for studying bird populations. As a result, bird vocalization analysis has become an active area of research.

1.3 Challenges in Manual Identification

Manual identification of bird species from audio recordings can be difficult and time-consuming. It requires trained experts who can distinguish subtle differences among bird calls. Environmental noise, overlapping sounds, and large datasets further complicate the process [13]. Human-based analysis may also introduce inconsistencies and errors. These challenges motivate the development of automated bird call recognition systems.

1.4 Role of Digital Signal Processing

Digital Signal Processing (DSP) provides techniques for analyzing and interpreting acoustic signals. DSP methods enable the extraction of meaningful information from raw audio recordings. Features such as frequency content, energy distribution, and temporal characteristics can be computed efficiently [14]. These features reveal important patterns associated with different bird species. Consequently, DSP forms the foundation of modern bioacoustic analysis systems.

1.5 Feature Extraction Techniques

Feature extraction transforms complex audio signals into compact numerical representations. Acoustic features capture both spectral and temporal properties of bird vocalizations. Common features include spectral centroid, spectral spread, entropy, spectral flux, and dominant frequency [15]. Effective feature extraction improves classification accuracy and reduces computational complexity. Therefore, it is considered a critical stage in automated recognition systems.

1.6 Machine Learning for Classification

Machine learning algorithms have revolutionized pattern recognition and classification tasks. These algorithms learn relationships between input features and output classes using training data. In bird call identification, machine learning models can automatically recognize species from acoustic characteristics [16]. They eliminate the need for manually defined classification rules. As a result, machine learning has become widely adopted in bioacoustic applications.

1.7 K-Nearest Neighbor Algorithm

The K-Nearest Neighbor (KNN) algorithm is a simple yet powerful supervised learning technique. It classifies unknown samples based on the labels of neighboring training samples in feature space. KNN is particularly effective for datasets with well-separated classes and low computational requirements. The algorithm is easy to implement and interpret [17]. These advantages make it suitable for bird call classification applications.

1.8 MATLAB-Based System Development

MATLAB provides a versatile platform for signal processing, visualization, and machine learning development. Its extensive libraries simplify the implementation of acoustic analysis algorithms. In this study, MATLAB is used to generate bird call signals, extract features, and train classification models [18]. Multiple visualization tools are employed to evaluate system performance. This integrated environment supports rapid development and testing.

1.9 Objectives of the Present Study

The primary objective of this work is to develop an automated bird call identification system using machine learning techniques. The study focuses on extracting discriminative acoustic features from bird vocalizations and classifying them accurately. Synthetic bird call datasets representing different species are utilized for training and testing [19]. Performance is evaluated using graphical and statistical measures. The goal is to demonstrate an efficient framework for species recognition.

1.10 Significance and Contributions

The proposed system contributes to the field of bioacoustics by combining signal processing and machine learning for automated bird monitoring. It provides a computationally efficient solution for species identification from audio signals [20]. The framework can support wildlife conservation, ecological research, and environmental assessment programs. Furthermore, it offers a scalable foundation for future studies involving larger datasets and advanced learning models. The research highlights the potential of intelligent acoustic monitoring technologies in biodiversity management.

2. Problem Statement

Bird populations are increasingly affected by habitat loss, climate change, urbanization, and human activities, creating an urgent need for continuous and accurate wildlife monitoring. Traditional bird species identification relies heavily on expert observers and manual analysis of field recordings, making large-scale monitoring expensive, time-consuming, and difficult to maintain over long periods. Many environments contain overlapping sounds, background noise, and varying recording conditions that further complicate reliable identification. Existing manual methods also introduce subjectivity, as recognition accuracy depends on the experience and availability of specialists. Although acoustic monitoring systems can collect vast amounts of audio data, converting these recordings into meaningful species information remains a major challenge. The key problem is the lack of an efficient, automated, and scalable system that can accurately identify bird species from their vocalizations. This research addresses that gap by combining digital signal processing and machine learning techniques to extract discriminative acoustic features and classify bird calls automatically. The proposed MATLAB-based framework aims to reduce human effort, improve consistency, and provide rapid species recognition suitable for ecological monitoring and conservation applications.

You can download the Project files here: Download files now. (You must be logged in).

3. Mathematical Approach

The proposed Bird Call Identification system utilizes digital signal processing and machine learning techniques to recognize bird species from acoustic signals. Initially, the recorded bird call signal is represented as a discrete-time sequence and transformed into the frequency domain using the Fast Fourier Transform (FFT). The FFT enables the extraction of important spectral characteristics that distinguish different bird vocalizations. Among the extracted features, the spectral centroid is used to measure the center of gravity of the frequency spectrum Centroid [21] and is calculated as:

• (C_s) = Spectral centroid (Hz)
• (f_k) = Frequency corresponding to the k-th spectral bin
• (X(k)) = Magnitude of the FFT at frequency bin k
• (N) = Total number of frequency bins
• (k) = Frequency bin index

The spectral centroid provides information about the brightness and dominant frequency distribution of a bird call. Another important feature is spectral entropy, which quantifies the randomness and complexity of the frequency spectrum. Higher entropy values indicate more complex acoustic patterns, while lower values represent tonal sounds. Spectral entropy [22] is computed as:

• (H) = Spectral entropy
• (P(k)) = Normalized spectral probability distribution
• (N) = Total number of frequency bins
• (k) = Frequency bin index
• (log_2) = Base-2 logarithmic function

After feature extraction, classification is performed using the K-Nearest Neighbor (KNN) algorithm. The Euclidean distance [23] metric is employed to measure similarity between the unknown bird call feature vector and training samples. The distance between two feature vectors is calculated as:

• (D(x,y)) = Euclidean distance between two feature vectors
• (x_i) = i-th feature of the unknown bird call
• (y_i) = i-th feature of the training sample
• (m) = Total number of extracted features
• (i) = Feature index

The classifier assigns the unknown bird call to the species most frequently represented among its nearest neighbors. Feature normalization is applied before classification to ensure equal contribution of all extracted parameters. The mathematical framework combines frequency-domain analysis, statistical feature extraction, and distance-based machine learning classification to achieve reliable bird species recognition. This approach improves classification accuracy while maintaining computational efficiency, making it suitable for real-time bioacoustic monitoring and wildlife conservation applications.

4. Methodology

The methodology of the proposed Bird Call Identification system consists of data generation, signal preprocessing, feature extraction, machine learning classification, and performance evaluation stages. Initially, a synthetic bird call database is created in MATLAB to represent different bird species, including Sparrow, Robin, Crow, and Canary. Each bird species is modeled using distinct acoustic patterns such as sinusoidal signals, chirp signals, and sawtooth waveforms to mimic realistic vocalizations. To simulate natural environmental conditions, an exponential decay envelope and additive Gaussian noise are incorporated into each generated signal [24]. The generated bird calls are then stored and labeled according to their corresponding species classes. In the preprocessing stage, the audio signals are standardized and prepared for analysis. Frequency-domain analysis is performed using the Fast Fourier Transform (FFT) to obtain spectral information from the bird calls. Subsequently, multiple acoustic features are extracted from each signal, including spectral centroid, spectral spread, spectral entropy, spectral flux, spectral roll-off frequency, zero-crossing rate, signal energy, dominant frequency, spectral flatness, and RMS value. These features form a multidimensional feature vector that characterizes the unique properties of each bird species. The complete feature dataset is then divided into training and testing subsets using a hold-out validation strategy, where 70% of the samples are used for training and 30% for testing. Feature normalization is applied to remove scale differences and improve classification performance. A K-Nearest Neighbor (KNN) classifier with five nearest neighbors and Euclidean distance metric is trained using the normalized training dataset [25]. The trained model learns the relationship between extracted acoustic features and bird species labels. During the testing phase, unknown bird calls are processed through the same feature extraction and normalization procedures before classification. The classifier predicts the most probable bird species based on feature similarity. Finally, system performance is evaluated using classification accuracy, confusion matrix analysis, waveform visualization, spectrogram analysis, frequency spectrum examination, acoustic feature representation, and feature-space clustering. The overall methodology provides an efficient framework for automated bird species recognition using acoustic signal processing and machine learning techniques.

5. Design Matlab Simulation and Analysis

The MATLAB simulation implements an automated Bird Call Identification system based on machine learning and acoustic feature analysis.

Table 1: Simulation Parameters

Table 1 represents the simulation begins by defining the sampling frequency, signal duration, and time vector required for generating bird vocalization signals. Four bird species, namely Sparrow, Robin, Crow, and Canary, are modeled using different waveform structures to emulate distinct acoustic characteristics. Synthetic bird calls are generated using sinusoidal, chirp, and sawtooth functions with random frequency variations to introduce natural diversity among samples. An exponential decay envelope is applied to each signal to mimic the fading behavior of real bird vocalizations. Gaussian noise is added to simulate environmental disturbances commonly present in field recordings. For every generated bird call, a feature extraction function is executed to compute important acoustic descriptors. These descriptors include spectral centroid, spectral spread, spectral entropy, spectral flux, spectral roll-off frequency, zero-crossing rate, signal energy, dominant frequency, spectral flatness, and RMS value. The extracted features are stored in a feature matrix, while corresponding species labels are assigned for supervised learning. The dataset is then divided into training and testing subsets using a 70:30 hold-out validation strategy. Feature normalization is performed to ensure that all extracted parameters contribute equally during classification. A K-Nearest Neighbor (KNN) classifier with five nearest neighbors and Euclidean distance metric is trained using the normalized training data. An unknown bird call is subsequently generated and processed through the same feature extraction and normalization procedures. The trained classifier predicts the bird species based on the similarity between the unknown feature vector and the training samples. The simulation produces six graphical outputs to visualize system behavior and performance. These outputs include the bird call waveform, spectrogram, frequency spectrum, acoustic feature vector, feature-space clustering plot, and confusion matrix. The waveform plot illustrates the time-domain structure of the bird call, while the spectrogram reveals its time-frequency characteristics. The frequency spectrum highlights dominant frequency components, and the feature clustering plot demonstrates the separability of different bird species in feature space. Finally, the confusion matrix evaluates classification accuracy and confirms the effectiveness of the proposed bird call recognition framework.

Figure 2: Bird Call Waveform

Figure 2 presents the time-domain waveform of the unknown bird call signal used for classification. The plot shows variations in signal amplitude with respect to time and illustrates the overall temporal behavior of the bird vocalization. The exponential decay envelope causes the signal amplitude to gradually decrease as time progresses, resembling natural bird calls. Random noise components added to the signal create realistic environmental conditions. This figure provides an initial understanding of the acoustic structure of the recorded bird sound.

Figure 3: Bird Call Spectrogram

Figure 3 displays the spectrogram of the unknown bird call, representing how frequency content changes over time. The horizontal axis corresponds to time, while the vertical axis represents frequency. Different color intensities indicate variations in signal energy across frequency bands. The chirp-based bird call produces a frequency sweep that is clearly visible in the spectrogram. This visualization helps identify unique time-frequency patterns useful for species recognition.

Figure 4: Frequency Spectrum

You can download the Project files here: Download files now. (You must be logged in).

Figure 4 illustrates the frequency-domain representation of the bird call obtained using the Fast Fourier Transform (FFT). The spectrum reveals the distribution of signal energy across different frequency components. Peaks in the graph indicate dominant frequencies that characterize the bird vocalization. The frequency spectrum assists in distinguishing bird species based on their spectral signatures. It also provides important information for extracting acoustic features used in classification.

Figure 5: MFCC-Like Acoustic Feature Vector

Figure 5 shows the extracted acoustic feature vector used as input to the machine learning classifier. Each stem corresponds to a specific feature such as spectral centroid, entropy, energy, or dominant frequency. The normalized feature values provide a compact numerical representation of the bird call. These features capture important spectral and temporal properties of the signal. The figure demonstrates how complex audio data can be transformed into meaningful classification parameters.

Figure 6: Bird Species Feature Clusters

Figure 6 presents the clustering of bird species in the feature space using spectral centroid and spectral spread as representative features. Each data point corresponds to a bird call sample, while different colors represent different bird species. Distinct clusters indicate that the extracted features effectively separate the species classes. Minimal overlap among clusters suggests strong discriminative capability of the feature extraction process. This figure validates the suitability of the selected features for machine learning classification.

Figure 7: Confusion Matrix

You can download the Project files here: Download files now. (You must be logged in).

Figure 7 presents the confusion matrix generated from the KNN classification results. The matrix compares actual bird species labels with predicted labels obtained from the classifier. Diagonal elements represent correctly classified samples, while off-diagonal elements indicate misclassifications. A higher concentration of values along the diagonal signifies better classification performance and higher accuracy. This figure provides a comprehensive assessment of the effectiveness and reliability of the bird call identification system.

6. Results and Discussion

The results obtained from the MATLAB simulation demonstrate the effectiveness of the proposed Bird Call Identification system in recognizing bird species using acoustic feature analysis and machine learning techniques. A dataset consisting of synthetic bird vocalizations representing Sparrow, Robin, Crow, and Canary species was successfully generated and processed. The extracted acoustic features captured significant spectral and temporal characteristics of the bird calls, enabling effective discrimination among species [26]. The waveform analysis revealed realistic signal behavior with amplitude decay and environmental noise effects similar to natural bird vocalizations. Spectrogram results clearly illustrated the time-frequency structures of bird calls, allowing distinctive frequency patterns to be observed for each species. Frequency spectrum analysis identified dominant spectral components that played a major role in species differentiation. The extracted feature vectors provided compact numerical representations of the acoustic signals and significantly reduced data complexity while preserving important information. Feature-space clustering showed clear separation among bird species, indicating that the selected acoustic features possess strong discriminative capability. Minimal overlap between clusters demonstrated the robustness of the feature extraction process [27]. The K-Nearest Neighbor classifier successfully learned the relationships between acoustic features and species labels during training. Testing results showed high classification accuracy, confirming the suitability of KNN for bird call recognition tasks. The confusion matrix revealed that most samples were correctly classified, with only a small number of misclassifications occurring between acoustically similar species. Feature normalization contributed to improved classifier stability and ensured balanced feature contributions during training and testing. The use of spectral centroid, spectral spread, entropy, dominant frequency, and energy features significantly enhanced recognition performance [28]. The system exhibited strong resistance to moderate noise levels introduced during signal generation. Computational complexity remained relatively low, making the framework suitable for real-time implementation. The generated visualizations provided valuable insights into signal behavior, feature distribution, and classification performance. Overall, the results validate the effectiveness of integrating digital signal processing with machine learning for automated bird species identification. The proposed framework offers a reliable, scalable, and computationally efficient solution for bioacoustic monitoring, wildlife conservation, and ecological research applications.

7. Conclusion

This study presented a Bird Call Identification system developed in MATLAB using acoustic feature extraction and machine learning techniques. Synthetic bird vocalizations representing multiple bird species were generated and analyzed to evaluate the effectiveness of the proposed framework. Important acoustic features, including spectral centroid, spectral spread, entropy, energy, dominant frequency, spectral flatness, and RMS value, were successfully extracted from bird calls. A K-Nearest Neighbor (KNN) classifier was employed to perform species recognition based on the extracted feature vectors [29]. Simulation results demonstrated high classification accuracy and effective separation of bird species within the feature space. Waveform, spectrogram, frequency spectrum, feature clustering, and confusion matrix analyses confirmed the reliability of the developed system. The integration of digital signal processing and machine learning enabled efficient and automated bird call recognition. The proposed methodology reduced the need for manual identification while maintaining robust classification performance. Furthermore, the framework is computationally efficient and suitable for real-time bioacoustic monitoring applications [30]. Overall, the developed Bird Call Identification system provides a practical and scalable solution for wildlife conservation, biodiversity assessment, and ecological research, while offering a strong foundation for future enhancements using advanced deep learning techniques and real-world audio datasets.

8. References

[1] L. R. Rabiner and B.-H. Juang, Fundamentals of Speech Recognition. Englewood Cliffs, NJ, USA: Prentice Hall, 1993.

[2] S. Mallat, A Wavelet Tour of Signal Processing: The Sparse Way, 3rd ed. Burlington, MA, USA: Academic Press, 2009.

[3] A. V. Oppenheim and R. W. Schafer, Discrete-Time Signal Processing, 3rd ed. Upper Saddle River, NJ, USA: Pearson, 2010.

[4] C. M. Bishop, Pattern Recognition and Machine Learning. New York, NY, USA: Springer, 2006.

[5] T. M. Cover and P. E. Hart, “Nearest neighbor pattern classification,” IEEE Trans. Inf. Theory, vol. 13, no. 1, pp. 21–27, Jan. 1967.

[6] K. Fukunaga, Introduction to Statistical Pattern Recognition, 2nd ed. San Diego, CA, USA: Academic Press, 1990.

[7] R. O. Duda, P. E. Hart, and D. G. Stork, Pattern Classification, 2nd ed. New York, NY, USA: Wiley, 2001.

[8] S. Theodoridis and K. Koutroumbas, Pattern Recognition, 4th ed. Burlington, MA, USA: Academic Press, 2009.

[9] I. Guyon and A. Elisseeff, “An introduction to variable and feature selection,” J. Mach. Learn. Res., vol. 3, pp. 1157–1182, 2003.

[10] T. Fawcett, “An introduction to ROC analysis,” Pattern Recognit. Lett., vol. 27, no. 8, pp. 861–874, 2006.

[11] J. Foote, “Content-based retrieval of music and audio,” in Proc. SPIE Multimedia Storage and Archiving Systems II, 1997, pp. 138–147.

[12] M. Slaney, “Auditory toolbox version 2,” Interval Research Corporation, Palo Alto, CA, USA, Tech. Rep. 1998.

[13] D. Ellis, “PLP and RASTA (and MFCC, and inversion) in MATLAB,” Columbia University, New York, NY, USA, Tech. Rep., 2005.

[14] B. C. Moore, An Introduction to the Psychology of Hearing, 6th ed. Leiden, The Netherlands: Brill, 2012.

[15] J. W. Cooley and J. W. Tukey, “An algorithm for the machine calculation of complex Fourier series,” Math. Comput., vol. 19, no. 90, pp. 297–301, 1965.

[16] D. Reynolds, “Gaussian mixture models,” in Encyclopedia of Biometrics. Boston, MA, USA: Springer, 2009, pp. 659–663.

[17] T. Hastie, R. Tibshirani, and J. Friedman, The Elements of Statistical Learning, 2nd ed. New York, NY, USA: Springer, 2009.

[18] S. Haykin, Neural Networks and Learning Machines, 3rd ed. Upper Saddle River, NJ, USA: Pearson, 2009.

[19] K. P. Murphy, Machine Learning: A Probabilistic Perspective. Cambridge, MA, USA: MIT Press, 2012.

[20] X. Huang, A. Acero, and H.-W. Hon, Spoken Language Processing. Upper Saddle River, NJ, USA: Prentice Hall, 2001.

[21] S. Mallat, A Wavelet Tour of Signal Processing, 3rd ed. Academic Press, 2009.

[22] L. Rabiner and B.-H. Juang, Fundamentals of Speech Recognition. Prentice Hall, 1993.

[23] C. M. Bishop, Pattern Recognition and Machine Learning. Springer, 2006.

[24] K. J. Piczak, “Environmental sound classification with convolutional neural networks,” in Proc. IEEE Int. Workshop Machine Learning for Signal Processing (MLSP), 2015, pp. 1–6.

[25] D. Stowell, D. Clayton, and M. D. Plumbley, “Birdsong recognition in complex acoustic environments,” Proc. Int. Soc. Music Inf. Retrieval Conf., pp. 321–326, 2013.

[26] H. Glotin, C. Artieres, and S. Vellinga, “LifeCLEF bird identification task overview,” in Working Notes of CLEF Conference, 2016, pp. 1–15.

[27] R. Bardeli, D. Wolff, F. Kurth, M. Koch, K.-H. Tauchert, and K.-H. Frommolt, “Detecting bird sounds in a complex acoustic environment and application to bioacoustic monitoring,” Pattern Recognit. Lett., vol. 31, no. 12, pp. 1524–1534, 2010.

[28] A. Acevedo, C. Corrada-Bravo, H. Villanueva-Rivera, and T. M. Aide, “Automated classification of bird and amphibian calls using machine learning,” J. Acoust. Soc. Am., vol. 123, no. 5, pp. 3059–3069, 2009.

[29] M. Somervuo, A. Harma, and S. Fagerlund, “Parametric representations of bird sounds for automatic species recognition,” IEEE Trans. Audio Speech Lang. Process., vol. 14, no. 6, pp. 2252–2263, Nov. 2006.

[30] S. Fagerlund, “Bird species recognition using support vector machines,” EURASIP J. Adv. Signal Process., vol. 2007, no. 1, pp. 1–8, 2007.

You can download the Project files here: Download files now. (You must be logged in).