- Research Article
- Open Access
A Decision-Tree-Based Algorithm for Speech/Music Classification and Segmentation
© Y. Lavner and D. Ruinskiy. 2009
- Received: 10 September 2008
- Accepted: 27 February 2009
- Published: 17 June 2009
We present an efficient algorithm for segmentation of audio signals into speech or music. The central motivation to our study is consumer audio applications, where various real-time enhancements are often applied. The algorithm consists of a learning phase and a classification phase. In the learning phase, predefined training data is used for computing various time-domain and frequency-domain features, for speech and music signals separately, and estimating the optimal speech/music thresholds, based on the probability density functions of the features. An automatic procedure is employed to select the best features for separation. In the test phase, initial classification is performed for each segment of the audio signal, using a three-stage sieve-like approach, applying both Bayesian and rule-based methods. To avoid erroneous rapid alternations in the classification, a smoothing technique is applied, averaging the decision on each segment with past segment decisions. Extensive evaluation of the algorithm, on a database of more than 12 hours of speech and more than 22 hours of music showed correct identification rates of 99.4% and 97.8%, respectively, and quick adjustment to alternating speech/music sections. In addition to its accuracy and robustness, the algorithm can be easily adapted to different audio types, and is suitable for real-time operation.
- Learning Phase
- Audio Signal
- Speech Segment
- Separation Power
- Music Signal
In the past decade a vast amount of multimedia data, such as text, images, video, and audio has become available. Efficient organization and manipulation of this data are required for many tasks, such as data classification for storage or navigation, differential processing according to content, searching for specific information, and many others.
A large portion of the data is audio, from resources such as broadcasting channels, databases, internet streams, and commercial CDs. To answer the fast-growing demands for handling the data, a new field of research, known as audio content analysis (ACA), or machine listening, has recently emerged, with the purpose of analyzing the audio data and extracting the content information directly from the acoustic signal  to the point of creating a "Table of Contents" .
Audio data (e.g., from broadcasting) often contains alternating sections of different types, such as speech and music. Thus, one of the fundamental tasks in manipulating such data is speech/music discrimination and segmentation, which is often the first step in processing the data. Such preprocessing is desirable for applications requiring accurate demarcation of speech, for instance automatic transcription of broadcast news, speech and speaker recognition, word or phrase spotting, and so forth. Similarly, it is useful in applications where attention is given to music, for example, genre-based or mood-based classification.
Speech/music classification is also important for applications that apply differential processing to audio data, such as content-based audio coding and compressing or automatic equalization of speech and music. Finally, it can also serve for indexing other data, for example, classification of video content through the accompanying audio.
One of the challenges in speech/music discrimination is characterization of the music signal. Speech is composed from a selection of fairly typical sounds and as such, can be represented well by relatively simple models. On the other hand, the assortment of sounds in music is much broader and produced by a variety of instruments, often by many simultaneous sources. As such, construction of a model to accurately represent and encompass all kinds of music is very complicated. This is one of the reasons that most of the algorithmic solutions developed for speech/music discrimination are practically adapted to the specific application they serve. A single comprehensive solution that will work in all situations is difficult to achieve. The difficulty of the task is increased by the fact that on many occasions speech is superimposed on the music parts, or vice versa.
1.1. Former Studies
Summary of Former studies.
Saunders, 1996 
Automatic real-time FM radio monitoring
Short-time energy, statistical parameters of the ZCR
Multivariate Gaussian classifier
Talk, commercials, music (different types)
Scheirer and Slaney, 1997 
Speech/music discrimination for automatic speech recognition
13 temporal, spectral and cepstral features (e.g., 4 Hz modulation energy, % of low energy frames, spectral rolloff, spectral centroid, spectral flux, ZCR, cepstrum-based feature, "rhythmicness"), variance of features across 1 sec.
Gaussian mixture model (GMM), K nearest neighbour (KNN), K-D trees, multidimensional Gaussian MAP estimator
FM radio (40 min): male and female speech, various conditions, different genres of music (training: 36 min, testing: 4 min)
94.2% (frame-by-frame), 98.6% (2.4 sec segments)
Foote, 1997 
Retrieving audio documents by acoustic similarity
12 MFCC, Short-time energy
Template matching of histograms, created using a tree-based vector quantizer, trained to maximize mutual information
409 sounds and 255 (7 sec long) clips of music.
No specific accuracy rates are provided. High rate of success in retrieving simple sounds.
Liu et al., 1997 
Analysis of audio for scene classification of TV programs
Silence ratio, volume std, volume dynamic range, 4 Hz freq, mean and std of pitch difference, speech, noise ratios, freq. centroid, bandwidth, energy in 4 sub-bands
A neural network using the one-class-in-one-network (OCON) structure
70 audio clips from TV programs (1 sec. long) for each scene class (training: 50, testing: 20)
Recognition of some of the classes is successful
Zhang and Kuo, 1999 
Audio segmentation/retrieval for video scene classification, indexing of raw audio visual recordings, database browsing
Features based on short-time energy, average ZCR, short-time fundamental frequency
A rule-based heuristic procedure for the coarse stage, HMM for the second stage
Coarse stage: speech, music, env. sounds and silence. Second stage: fine-class classification of env. sounds.
90% (coarse stage)
Williams and Ellis, 1999 
Segmentation of speech versus nonspeech in automatic speech recognition tasks
Mean per-frame entropy and average probability "dynamism", background-label energy ratio, phone distribution match—all derived from posterior probabilities of phones in hybrid connectionist-HMM framework
Gaussian likelihood ratio test
Radio recordings, speech (80 segments, 15 sec. each) and music (80, 15), respectively. Training: 75%, testing: 25%.
100% accuracy with 15 seconds long segments 98.7% accuracy with 2.5-seconds long segments
El-Maleh et al., 2000 
Automatic coding and content-based audio/video retrieval
LSF, differential LSF, measures based on the ZCR of high-pass filtered signal
KNN classifier and quadratic Gaussian classifier (QCG)
Several speakers, different genres of music (training: 9.3 min. and 10.7 min., resp.)
Frame level (20 ms): music 72.7% (QGC), 79.2% (KNN). Speech 74.3% (QGC), 82.5% (KNN). Segment level (1 sec.), music 94%–100%, speech 80%–94%.
Buggati et al., 2002 
"Table of Content description" of a multi-media document
ZCR-based features, spectral flux, short-time energy, cepstrum coefficients, spectral centroids, ratio of the high-frequency power spectrum, a measure based on syllabic frequency
Multivariate Gaussian classifier, neural network (MLP)
30 minutes of alternating sections of music and speech (5 min each)
95%–96% (NN). Total error rate: 17.7% (Bayesian classifier), 6.0% (NN).
Lu, Zhang, and Jiang, 2002 
Audio content analysis in video parsing
High zero-crossing rate ratio (HZCRR), low short-time energy ratio (LSTER), linear spectral pairs, band periodicity, noise-frame ratio (NFR)
3-step classification: 1. KNN and linear spectral pairs-vector quantization (LSP-VQ) for speech/nonspeech discrimination. 2. Heuristic rules for nonspeech classification into music/background noise/silence. 3. Speaker segmentation
MPEG-7 test data set, TV news, movie/audio clips. Speech: studio recordings, 4 kHz and 8 kHz bandwidths, music: songs, pop (training: 2 hours, testing: 4 hours).
Speech 97.5%, music 93.0%, env. sound 84.4%. Results of only speech/music discrimination: 98.0%
Ajmera et al., 2003 
Automatic transcription of broadcast news
Averaged entropy measure and "dynamism" estimated at the output of a multilayer perceptron (MLP) trained to emit posterior probabilities of phones. MLP input: 13 first cepstra of a 12th-order perceptual linear prediction filter.
2-state HMM with minimum duration constraints (threshold-free, unsupervised, no training).
4 files (10 min. each): alternate segments of speech and music, speech/music interleaved
GMM: Speech 98.8%, Music 93.9%. Alternating, variable length segments (MLP): Speech 98.6%, Music 94.6%.
Burred and Lerch, 2004 
Audio classification (speech/ music/back- ground noise), music classification into genres
Statistical measures of short-time frame features: ZCR, spectral centroid/rolloff/flux, first 5 MFCCs, audio spectrum centroid/flatness, harmonic ratio, beat strength, rhythmic regularity, RMS energy, time envelope, low energy rate, loudness, others
KNN classifier, 3-component GMM classifier
3 classes of speech, 13 genres of music and background noise: 50 examples for each class (30 sec each), from CDs, MP3, and radio.
94.6% /96.3% (hierarchical approach and direct approach, resp.)
Barbedo and Lopes, 2006 
Automatic segmentation for real-time applications
Features based on ZCR, spectral rolloff, loudness and fundamental frequencies
KNN, self-organizing maps, MLP neural networks, linear combinations
Speech (5 different conditions) and music (various genres)more than 20 hours of audio data, from CDs, Internet radio streams, radio broadcasting, and coded files.
Noisy speech 99.4%, Clean speech 100%, Music 98.8%, Music without rap 99.2%. Rapid alternations: speech 94.5%, music 93.2%.
Muñoz-Expósito et al., 2006 
Intelligent audio coding system
Warped LPC-based spectral centroid
3-component GMM, with or without fuzzy rules-based system
Speech (radio and TV news, movie dialogs, different conditions); music (various genres, different instruments/singers) -1 hour for each class.
GMM: speech 95.1%, music 80.3%. GMM with fuzzy system: speech 94.2%, music 93.1%.
Alexandre et al, 2006 
Speech/music classification for musical genre classification
Spectral centroid/rolloff, ZCR, short-time energy, low short-time energy ratio (LSTER), MFCC, voice-to-white
Fisher linear discriminant, K nearest-neighbour
Speech (without background music), and music without vocals. (training: 45 min, testing: 15 min)
Music 99.1%, speech 96.6%. Individual features: 95.9% (MFCC), 95.1% (voice to white).
The major differences between the different studies lie in the exact classification algorithm, even though some popular classifiers (K-nearest neighbour, Gaussian multivariate, neural network) are often used as a basis. Finally, in each study, different databases are used for training and testing the algorithm. It is worth noting that in most of the studies, especially the early ones, these databases are fairly small [6, 7]. Only in a few works large databases are used [8, 9].
1.2. The Algorithm
In this paper we present an efficient algorithm for segmentation of audio signals into speech or music. The central motivation to our study is consumer audio applications, where various real-time enhancements are often applied to music. These include differential frequency gain (equalizers) or spatial effects (such as simulation of surround and reverberation). While these manipulations can improve the perceptive quality of music, applying them to speech can cause distortions (for instance, bass amplification can cause an unpleasant booming effect).
As many audio sources, such as radio broadcasting streams, live performances, or movies, often contain sections of pure speech mixed between musical segments, an automatic real-time speech/music discrimination system may be used to allow the enhancement of music without introducing distortions to the speech.
Considering the application at hand, our algorithm aims to achieve the following:
(i)Pure speech must be identified correctly with very high accuracy, to avoid distortions when enhancements are applied.
(ii)Songs that contain a strong instrumental component together with voice should be classified as music, just like purely instrumental tracks.
(iii)Audio that is neither speech, nor music (noise, environmental sounds, silence, and so forth) can be ignored by the classifier, as it is not important for the application of the manipulations. We can therefore assume that a priori the audio belongs to one of the two classes.
(iv)The algorithm must be able to operate in real time with a low computational cost and a short delay.
The algorithm proposed here answers all these requirements: on one hand, it is highly accurate and robust, and on the other hand, simple, efficient, and adequate for real-time implementation. It achieves excellent results in minimizing misdetection of speech, due to a combination of the feature choice and the decision tree. The percentage of correct detection of music is also very high. Overall the results we obtained were comparable to the best of the published studies, with a confidence level higher than most, due to the large size of test database used.
The algorithm uses various time domain parameters of the audio signal, such as the energy, zero-crossing rate, and autocorrelation as well as frequency domain parameters (spectral energy, MFCC, and others).
The algorithm consists of two stages. The first stage is a supervised learning phase, based on a statistical approach. In this phase training data is collected from speech and music signals separately, and after processing and feature extraction, optimal separation thresholds between speech and music are set for each analyzed feature separately.
In the second stage, the processing phase, an input audio signal is divided into short-time segments and feature extraction is performed for each segment. The features are then compared to their corresponding thresholds, which were set in the learning phase, and initial classification of the segment as speech or music is carried out. Various post-decision techniques are applied to improve the robustness of the classification.
Our test database consisted of 12+ hours of speech and 20+ hours of music. This database is significantly larger than those used for testing in the majority of the aforementioned studies. Tested on this database, the algorithm proved to be highly accurate both in the correctness of the classification and the segmentation accuracy. The processing phase can also be applied in a real-time environment, due to low computation load of the process, and the fact that the classification is localized (i.e., a segment is classified as speech or music independently of other segments). A commercial product based on the proposed algorithm is currently being developed by Waves Audio, and a provisional patent has been filed.
The rest of the paper is arranged as follows: in Section 2 we describe the learning procedure, during which the algorithm is "trained" to distinguish between speech and music, as well as the features used for the distinction. Next, in Section 3, the processing phase and the classification algorithm are described. Section 4 provides evaluation of the algorithm in terms of classification success and comparison to other approaches, and is followed by a conclusion (Section 5).
2.1. Music and Speech Material
The music material for the training phase was derived mostly from CDs or from databases, using high bitrate signals with a total duration of 60 minutes. The material contained different genres and types of music such as classical music, rock and pop songs, folk music, etc.
The speech material was collected from free internet speech databases, also containing a total of 60 minutes. Both high and low bitrate signals were used.
2.2. General Algorithm
2.3. Computation of Features
Each of the speech signals and music signals in the learning phase is divided into consecutive analysis frames of length with hop size , where and are in samples, corresponding to 40 milliseconds and 20 milliseconds, respectively. For each frame, the following features are computed:
Band Energy Ratio
The band energy ratio captures the distribution of the spectral energy in different frequency bands. The spectral energy in a given band is defined as follows: Let denote one frame of the audio signal ( ), and let denote the Discrete Fourier Transform (DFT) of . The values of for correspond to discrete frequency bins from to , with indicating half of the sampling rate . Let denote the frequency in Hz. The DFT bin number corresponding to is given by
For a given frequency band the total spectral energy in the band is given by
Finally, if the spectral energies of the two bands and are denoted and , respectively, the ratio is computed on a logarithmic scale, as follows:
We used two features based on band energy ratio—the low energy ratio, defined as the ratio between the spectral energy below 70 Hz and the total energy, and the high energy ratio, defined as the ratio between the energy above 11 KHz and the total energy, where the sampling frequency is 44 KHz.
The autocorrelation coefficient is defined as the highest peak in the short-time autocorrelation sequence and is used to evaluate how close the audio signal is to a periodic one. First, the normalized autocorrelation sequence of the frame is computed:
Next, the highest peak of the autocorrelation sequence between and is located, where and correspond to periods between 3 milliseconds and 16 milliseconds (which is the expected fundamental frequency range in voiced speech). The autocorrelation coefficient is defined as the value of this peak:
Mel Frequency Cepstrum Coefficients
The mel frequency cepstrum coefficients (MFCCs) are known to be a compact and efficient representation of speech data [17, 18]. The MFCC computation starts by taking the DFT of the frame and multiplying it by a series of triangularly shaped ideal band-pass filters , where the central frequencies and widths of the filters are arranged according to the mel scale . Next, the total spectral energy contained in each filter is computed:
Finally, the MFCC sequence is obtained by computing the Discrete Cosine Transform (DCT) of the logarithm of the energy sequence :
We computed the first 10 MFC coefficients for each frame. Each individual MFC coefficient is considered a feature. In addition, the MFCC difference vector between neighboring frames is computed, and the Euclidean norm of that vector is used as an additional feature:
where represents the index of the frame.
Spectrum Rolloff Point
In our study is used.
namely, the sum of the squared frame-to-frame difference of the DFT magnitudes , where and are the frame indices.
2.4. Computation of Feature Statistics
Mean value and standard deviation of the feature across the segment.
(ii)Mean value and standard deviation of the difference magnitude between consecutive analysis points.
In addition to that, for the zero-crossing rate, the skewness (third central moment, divided by the cube of the standard deviation) and the skewness of the difference magnitude between consecutive analysis frames are also computed.
For the energy we also measure the low short time energy ratio (LSTER, ). The LSTER is defined as the percentage of frames within the segment whose energy level is below one third of the average energy level across the segment.
2.5. Threshold Setting and Probability Density Function Estimation
In the learning phase, training data is collected for speech segments and for music segments separately. For each feature and each statistical parameter the corresponding probability density functions (PDFs) are estimated—one for speech segments and one for music segments. The PDFs are computed using a nonparametric technique with a Gaussian kernel function for smoothing.
Extreme speech threshold—defined as the value beyond which there are only speech segments, that is, 0% error based on the learning data.
Extreme music threshold—same as 1, for music.
High probability speech threshold—defined as the point in the distribution where the difference between the height of the speech PDF and the height of the music PDF is maximal. This threshold is more permissive than the extreme speech threshold: values beyond this threshold are typically exhibited by speech, but a small error of music segments is usually present. If this error is small enough, and on the other hand a significant percentage of speech segments are beyond this threshold, the feature may be a good candidate for separation between speech and music.
High probability music threshold—same as 3, for music.
(5)Separation threshold—defined as the value that minimizes the joint decision error, assuming that the prior probabilities for speech and for music are equal.
inclusion fraction ( )—the percentage of correct segments that exceed the threshold (for the speech threshold this refers to speech segments, and for the music threshold this refers to music segments);
error fraction ( )—the percentage of incorrect segments that exceed the threshold. For speech thresholds these are the music segments, and for music thresholds these are the speech segments. Note that by the definition of the extreme thresholds, their error fractions are 0.
2.6. Feature Selection
With a total of over 20 features computed on the frame level and 4–6 statistical parameters computed per feature on the segment level, the feature space is quite large. More importantly, not all features contribute equally, and some features may be very good in certain aspects, and bad in others. For example, a specific feature may have a very high value of for the extreme speech threshold, but a very low value of for the extreme music threshold, making it suitable for one feature group, but not the other.
The main task here is to select the best features for each classification stage to ensure high discrimination accuracy and to reduce the dimension of the feature space.
The "usefulness" score is computed separately for each feature and each of the thresholds. The feature ranking method is different for each of the three threshold types (extreme speech/music, high probability speech/music, and separation).
For the extreme thresholds, the features are ranked according to the value of the corresponding inclusion fraction . When is large, the feature is likely to be more useful for identifying typical speech (resp., music) frames.
let be the separation power (for the extreme thresholds we set , whereas for the high probability thresholds we use );
in the first stage select the first feature that maximizes : ;
the second feature is computed so that
where and are weighting factors (typically ), determining the relative contributions of and of the mutual correlation :
the th feature is computed using
For the separation threshold we originally tried the Fisher Discriminant Ratio (FDR, ) as a measure of the feature separation power:
where , are the mean values and , are the standard deviations, for speech and music, respectively. As in the first two stages, the features were selected according to a combination of the FDR and the mutual correlation.
A small improvement was achieved by using the sequential floating (forward-backward) selection procedure detailed in [20, 21]. The advantage of this procedure is that it considers separation power of entire feature vector as a whole, and not just as combination of individual features.
To measure the separation power of the feature vectors, we computed the scatter matrices and . is the within-class scatter matrix, defined as the normalized sum of the class covariance matrices and :
is the mixture scatter matrix, defined as the covariance matrix of the feature vector (all samples, both speech and music) around the global mean.
Finally, the separation criterion is defined as
This criterion tends to take large values when the within-class scatter is small, that is, the samples are well-clustered around the class mean, but the overall scatter is large, implying that the clusters are well-separated. More details can be found in .
2.7. Best Features for Discrimination
Best features for each of the five thresholds.
(1) 9th MFCC (mean val. of diff. mag.)
(2) Energy (std)
(3) 9th MFCC (std of diff. mag.)
(1) High Band Energy Ratio (mean value)
(2) Spectral rolloff point (mean value)
(3) Spectral centroid (mean value)
High probability speech
(1) Energy (std)
(2) 9th MFCC (mean val. of diff. mag.)
(3) Energy (mean val. of diff. mag.)
(4) Autocorrelation (std)
High probability music
(1) Energy (mean val. of diff. mag.)
(2) Energy (std)
(3) 9th MFCC (std of diff. mag.)
(4) Autocorrelation (std of diff. mag.)
(5) ZCR (skewness)
(6) ZCR (skewness of diff. mag.)
(1) Energy (std)
(2) Energy (mean val. of diff. mag.)
(3) Autocorrelation (std)
(4) 9th MFCC (std of diff. mag.)
(5) Energy (std of diff. mag.)
(6) 9th MFCC (mean val. of diff. mag.)
(7) 7th MFCC (mean val. of diff. mag.)
(8) 4th MFCC (std)
(9) 7th MFCC (std of diff. mag.)
(10) Autocorrelation (std of diff. mag.)
As can be seen from the table, certain features, for example, the energy, the autocorrelation, and the 9th MFCC are useful in multiple stages, while some, like the spectral rolloff point, are used only in one of the stages. Also it can be noticed that some of the features considered in the learning phase were found inefficient in practice and were eliminated from the features set in the test phase. As the procedure is automatic, the user does not even have to know which features are selected, and in fact very different sets of features were sometimes selected for different thresholds.
3.1. Streaming and Feature Computation
The input signal is divided into consecutive segments, with segment size . Each segment is further divided into consecutive and overlapping frames with frame size (as in the learning phase) and hop size , where typically . For each such frame, the features that were chosen by the feature selection procedure (Section 2.6) are computed. Consequently, the feature statistics are computed over the segment (of length ) and compared to the predefined thresholds, which were also set during the learning phase. This comparison is used as a basis for classification of the segment either as speech or as music, as described below.
In order to provide better tracking of the changes in the signal, the segment hop size , which represents the resolution of the decision, is set to a small fraction of the segment size (typically ).
For the evaluation tests (Section 4) the following values were used: milliseconds, milliseconds, seconds.
3.2. Initial Classification
The initial decision is carried out for each segment independently of other segments. In this decision each segment receives a grade between and 1, where positive grades indicate music and negative grades indicate speech, and the actual value represents the degree of certainty in the decision (e.g., means speech/music with high certainty).
No. of features in the extreme speech (music) set that surpass their thresholds
No. of features in the high probability speech (music) set that surpass their thresholds
No. of features in the separation set that are classified as speech (music)
A segment receives a grade of if one of the following takes place:
(i) and (i.e., at least one of the features is above its corresponding extreme speech threshold; whereas no feature surpasses the extreme or the high probability music thresholds);
(ii) and (we allow if is at least 2);
(iii) and , where , is the set of all features used with the high probability speech threshold (i.e., if a decision cannot be made using the extreme thresholds, we demand a large majority of the high probability thresholds to classify the segment as speech with high certainty).
The above combination of rules allows classifying a segment as speech in cases where its feature vector is located far inside the speech half-space along some of the feature axes, and at the same time, is not far inside the music half-space along any of the axes. In such cases we can be fairly certain of the classification. It is expected that if the analyzed segment is indeed speech, it will rarely exhibit any features above the extreme or high probability music thresholds.
Similarly, a segment gets a grade of (that is considered as music with high certainty) if one of the following takes place:
(i) and ,
(ii) and ,
(iii) , and , where , is the set of all features used with the high probability speech threshold.
If none of the above applies, the decision is based on the separation threshold as follows:
where is the set of features used with the separation threshold. Note that , and , so the received grade is always between and 1, and in some way reflects the certainty with which the segment can be classified as speech or music.
3.3. Smoothing and Final Classification
In most audio signals, speech-music and music-speech transitions are not very common (for instance, musical segments are usually at least one minute long, and typically several minutes or longer). When the classification of an individual segment is based solely on data collected from that segment (as described above), erroneous decisions may lead to classification results that alternate more rapidly than normally expected. To avoid this, the initial decision is smoothed by a weighted average with past decisions, using an exponentially decaying "forgetting factor," which gives more weight to recent segments:
where is the length of the averaging period, is the time constant, and is the normalizing constant. Alternatively, we tried a median filter for the smoothing. Both approaches achieved comparable results.
Following the smoothing procedure, discretization to a binary decision is performed as follows: a threshold value is determined. Values above or below are set to 1 or , respectively, whereas values between and are treated according to the current trend of , that is, if is on the rise, and otherwise, where is the binary desicion.
Additionally, a four-level decision is possible, where values in are treated as "weakly speech" or "weakly music." The four-level decision mode is useful for mixed content signals, which are difficult to firmly classify as speech or music.
To avoid erroneous transitions in long periods of either music or speech, we adapt the threshold over time as follows: let be the threshold at time , and be the binary decision values of the current and the previous time instants, respectively. We have the following:
where is a predefined multiplier, is the initial value of the threshold, and is a minimal value, which is set so that the threshold will not reach a value of zero. This mechanism ensures that whenever a prolonged music (or speech) period is processed, the absolute value of the threshold is slowly decreased towards the minimal value. When the decision is changed, the threshold value is reset to .
4.1. Speech and Music Material
The speech material for the evaluation was collected from various free internet speech databases, such as the International Dialects of English Archive (http://www.ku.edu/~idea/), the World Voices Collection (http://www.world-voices.com/), the Indian institute of scientific heritage (http://www.iish.org/), and others. The test set contained both clean speech and noisy speech. Most of the data were MP3 files at bitrates of 128Kbps and higher or uncompressed PCM WAV files.
The music material for the test set contained music of different genres, such as classical, folk, pop, rock, metal, new age, and others. The source material was either in cd audio format (uncompressed) or mp3 (64 kbps and higher).
4.2. Evaluation of Long Tracks
Total correct identification rate for music and for speech.
Breakdown of identification rates for different musical genres.
4.3. Evaluation of Alternating Sections
Correct identification rates for alternating sections.
4.01 (std = 0.51)
5.13 (std = 0.71)
Most of the sections were identified perfectly. Among the speech sections only 2 out of 40 sections were problematic, where in one case the speech was excerpted from a TV soundtrack with background music, and the other contained strong background noise. Among the music, only 3 out of 40 sections had errors, 2 of them containing a strong vocal element. The high success rates are even more satisfying considering the fact that most of the speech sections were with a background noise.
4.4. Gradual Transitions
4.5. Comparison with Other Approaches
In order to evaluate the algorithm in the context of existing work, we compared it to traditional approaches, such as the Support Vector Machine (SVM, [22, 23]) and the Linear Discriminant Analysis (LDA, ). For the training, our original training databases were used, and for the testing we chose the Scheirer-Slaney database, provided to us by Dan Ellis. The database consists of radio recordings, with 20 minutes of speech and 20 minutes of music. This allows us to compare our results directly with past studies that used the same database such as [6, 12, 16].
In our comparative tests, the optimal feature sets for each classifier (the proposed algorithm, SVM and LDA) were chosen using the automatic feature selection procedure detailed in Section 2.6. For the SVM the Gaussian (radial basis function) kernel was used, as it has some advantages compared to the linear SVM and others. The data for the SVM training was automatically scaled to zero mean and unit variance, and the free parameters were selected using grid search, and evaluated through cross-validation [24, 25].
The SVM and LDA algorithms were tested individually, as stand-alone classifiers, as well as in the general framework of our decision tree, replacing the separation stage of the initial classification (Section 3.2), that is, was assigned according to the output of the classifier, and the postprocessing techniques described in Section 3.3 were applied to it. Additionally, we tested a combination of the two techniques—with LDA being used for feature selection, and SVM for the classification.
4.6. The Importance of Feature Selection
Effect of number of features on performance of SVM classifier.
Number of features
Correct identifaction rate
In this paper we presented an algorithm for speech/music classification and segmentation, based on a multistage statistical approach. The algorithm consists of a training stage with an automatic selection of the most efficient features and the optimal thresholds, and a processing stage, during which an unknown audio signal is segmented into speech or music through a sieve-like rule-based procedure.
Tested on a large test set of about 34 hours of audio from different resources, and of varying quality, the algorithm demonstrated very high accuracy in the classification (98%, and over 99% under certain conditions), as well as fast response to transitions from music to speech and vice versa. The test database is among the largest used in similar studies and suggests that the algorithm is robust and applicable in various testing conditions.
The algorithm was tested against standard approaches such as SVM and LDA, and showed comparable, often slightly better, results. On the other hand, the decision tree uses simple heuristics, which are easy to implement in a variety of hardware and software frameworks. This together with its fast and highly localized operation makes it suitable for real-time systems and for consumer-grade audio processing applications. A commercial product based on the presented algorithm is currently being developed by Waves Audio, and a provisional patent has been filed.
Another advantage of the presented algorithm is the generic training procedure, which allows testing any number of additional features and selecting the optimal feature set using an automatic and flexible procedure.
One possible direction for improvement of the algorithm is increasing the segmentation accuracy in speech-music and music-speech transitions. In an offline application this could be achieved, without harming the overall identification rate, by postprocessing the boundary regions using smaller segments.
The authors would like to thank I. Neoran, Director of R&D of Waves Audio, for his valuable discussions and ideas, and Y. Yakir for technical assistance. They also thank D. Ellis for providing them the database they used in part of the experiments. Finally, they would like to thank the reviewers for many good remarks which helped them improve the paper.
- Burred JJ, Lerch A: Hierarchical automatic audio signal classification. Journal of the Audio Engineering Society 2004,52(7-8):724-739.Google Scholar
- Bugatti A, Flammini A, Migliorati P: Audio classification in speech and music: a comparison between a statistical and a neural approach. EURASIP Journal on Applied Signal Processing 2002,2002(4):372-378. 10.1155/S1110865702000720View ArticleMATHGoogle Scholar
- Muñoz-Expósito JE, Galán SG, Reyes NR, Candeas PV, Peña FR: A fuzzy rules-based speech/music discrimination approach for intelligent audio coding over the Internet. Proceedings of the 120th Audio Engineering Society Convention (AES '06), May 2006, Paris, France paper number 6676Google Scholar
- Saunders J: Real-time discrimination of broadcast speech/music. Proceedings of IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP '96), May 1996, Atlanta, Ga, USA 2: 993-996.Google Scholar
- Liu Z, Huang J, Wang Y, Chen IT: Audio feature extraction and analysis for scene classification. Proceedings of the 1st IEEE Workshop on Multimedia Signal Processing (MMSP '97), June 1997, Princeton, NJ, USA 343-348.Google Scholar
- Scheirer E, Slaney M: Construction and evaluation of a robust multifeature speech/music discriminator. Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP '97), April 1997, Munich, Germany 2: 1331-1334.Google Scholar
- Ajmera J, McCowan IA, Bourlard H: Robust HMM-based speech/music segmentation. Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP '03), May 2002, Orlando, Fla, USA 1: 297-300.Google Scholar
- Carey MJ, Parris ES, Lloyd-Thomas H: A comparison of features for speech, music discrimination. Proceedings of IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP '99), March 1999, Phoenix, Ariz, USA 1: 149-152.View ArticleGoogle Scholar
- Lu L, Zhang H-J, Jiang H: Content analysis for audio classification and segmentation. IEEE Transactions on Speech and Audio Processing 2002,10(7):504-516. 10.1109/TSA.2002.804546View ArticleGoogle Scholar
- Foote JT: A similarity measure for automatic audio classification. Proceedings of the AAAI Spring Symposium on Intelligent Integration and Use of Text, Image, Video, and Audio Corpora, March 1997, Stanford, Calif, USAGoogle Scholar
- Zhang T, Kuo C-CJ: Hierarchical classification of audio data for archiving and retrieving. Proceedings of IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP '99), March 1999, Phoenix, Ariz, USA 6: 3001-3004.View ArticleGoogle Scholar
- Williams G, Ellis DPW: Speech/music discrimination based on posterior probability features. Proceedings of the 6th European Conference on Speech Communication and Technology (EUROSPEECH '99), September 1999, Budapest, Hungary 687-690.Google Scholar
- El-Maleh K, Klein M, Petrucci G, Kabal P: Speech/music discrimination for multimedia applications. Proceedings of IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP '00), June 2000, Istanbul, Turkey 6: 2445-2448.Google Scholar
- Ajmera J, McCowan I, Bourlard H: Speech/music segmentation using entropy and dynamism features in a HMM classification framework. Speech Communication 2003,40(3):351-363. 10.1016/S0167-6393(02)00087-0View ArticleGoogle Scholar
- Barbedo JGA, Lopes A: A robust and computationally efficient speech/music discriminator. Journal of the Audio Engineering Society 2006,54(7-8):571-588.Google Scholar
- Alexandre E, Rosa M, Caudra L, Gil-Pita R: Application of Fisher linear discriminant analysis to speech/music classification. Proceedings of the 120th Audio Engineering Society Convention (AES '06), May 2006, Paris, France paper number 6678Google Scholar
- Davis SB, Mermelstein P: Comparison of parametric representations for monosyllabic word recognition in continuously spoken sentences. IEEE Transactions on Acoustics, Speech, and Signal Processing 1980,28(4):357-366. 10.1109/TASSP.1980.1163420View ArticleGoogle Scholar
- Quatieri TF: Discrete-Time Speech Signal Processing. Prentice-Hall, Englewood Cliffs, NJ, USA; 2001.Google Scholar
- Stevens SS, Volkman J: The relation of pitch to frequency: a revised scale. American Journal of Psychology 1940,53(3):329-353. 10.2307/1417526View ArticleGoogle Scholar
- Theodoridis S, Koutroumbas K: Pattern Recognition. 4th edition. Academic Press, New York, NY, USA; 2003.Google Scholar
- Pudil P, Novovicova J, Kittler J: Floating search methods in feature selection. Pattern Recognition Letters 1994,15(11):1119-1125. 10.1016/0167-8655(94)90127-9View ArticleGoogle Scholar
- Cristianini N, Shawe-Taylor J: An Introduction to Support Vector Machines and other Kernel-Based Learning Methods. Cambridge University Press, Cambridge, UK; 2000.View ArticleGoogle Scholar
- Burges CJC: Simplified support vector decision rules. Proceedings of the 13th International Conference on Machine Learning (ICML '96), July 1996, Bari, Italy 71-77.Google Scholar
- Staelin C: Parameter selection for support vector machines. HP Laboratories Israel, Haifa, Israel; 2003.Google Scholar
- Hsu CW, Chang CC, Lin CJ: A practical guide to support vector classification. National Taiwan University, Taipei, Taiwan; 2003.Google Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.